Character count determination for a digital image

[0001] This application is a continuation of U.S. patent application Ser. No. 15/014,918, filed on Feb. 3, 2016, which is a continuation of U.S. patent application Ser. No. 14/056,565 (now U.S. Pat. No. 9,286,514 issued on Mar. 15, 2016), filed on Oct. 17, 2013, the entirety of all of which are incorporated by reference herein.

[0002] 1. Technical Field

[0003] This disclosure relates to determining a character count for an image, such as a financial document image. This disclosure also relates to determining the character count of the image without recognizing the identity of the characters in the image.

[0004] 2. Description of Related Art

[0005] Systems may receive digital images for processing. As one example, an electronic device may capture an image of a financial document, such as a check. The user can submit the image of the check to a financial institution server for processing and deposit of the check. However, the check image may be degraded in multiple ways. The check may be overly cropped by the user such that important fields or portions of the check are cropped out of the check image. The image capture device of the electronic device may capture a blurry image of the check. These degradations may inhibit subsequent processing of the check image.

[0006] The descriptions below include methods, systems, logic, and devices for processing an image and determining a number of characters in the image with recognizing or attempting to recognize the actual text or characters in the image. In one aspect, a method is performed by circuitry, such as a processor, in an electronic device. The method performed by the circuitry includes receiving a financial document image and identifying a text chunk in the financial document image by determining a first pixel of the financial document image as the top left pixel of the text chunk based on a top left score of the first pixel and determining a second pixel of the financial document image as top right pixel of the text chunk based on a top right score of the second pixel. The method further includes determining a character count for the text chunk without recognizing any particular character in the text chunk.

[0007] In another aspect, a system includes a memory and a processor. The processor is operable to receive a financial document image and identify a text chunk in the financial document image by determining a first pixel of the financial document image as top left pixel of the text chunk based on a top left score of the first pixel and determining a second pixel of the financial document image as top right pixel of the text chunk based on a top right score of the second pixel. The processor is also operable to determine a chunk extension for the text chunk and add the chunk extension to the text chunk. After adding the chunk extension to the text chunk, the processor is operable to determine a character count for the text chunk without recognizing any particular character in the text chunk.

[0008] In another aspect, a non-transitory computer readable medium includes instructions that, when executed by a processor, cause the processor to receive a financial document image; identify an interest region in the financial document image; and identify a text chunk in the interest region of the financial document image. The instructions cause the processor to identify the text chunk by determining a first pixel of the financial document image as top left pixel of the text chunk based on a top left score of the first pixel and determining a second pixel of the financial document image as top right pixel of the text chunk based on a top right score of the second pixel. The instructions also cause the processor to determine a character count for the text chunk in the interest region of the financial document image without recognizing any particular character in the text chunk; determine a character count for the interest region by summing the character count for the text chunk with a character count of any additional text chunks in the interest region; and determine whether the character count for the interest region exceeds a minimum character count threshold specifically for the interest region.

[0021] FIG. 1 shows an example of a system 100 for determining a character count of an image. The system 100 in FIG. 1 includes an electronic device 102 communicatively linked to an image processing system 104 through a communication network 106. The electronic device 102 may include or communicate with an image capture device that captures digital images, such as a digital camera or scanning device. In FIG. 1, the electronic device 102 is a mobile communication device, e.g., a cellular telephone with a digital camera. However, the electronic device 102 may take any number of forms, including as examples a laptop, digital camera, personal digital assistant (PDA), tablet device, portable music player, desktop computer, any image scanning device, or others. The electronic device 102 may also include any number of communication interfaces supporting communication through the communication network 106.

[0022] The communication network 106 may include any number of networks for communicating data. In that regard, the communication network 106 may include intermediate network devices or logic operating according to any communication number of mediums, protocols, topologies, or standards. As examples, the communication network 106 may communicate across any of the following mediums, protocols, topologies and standards: Ethernet, cable, DSL, Multimedia over Coax Alliance, power line, Ethernet Passive Optical Network (EPON), Gigabit Passive Optical Network (GPON), any number of cellular standards (e.g., 2G, 3G, Universal Mobile Telecommunications System (UMTS), GSM® Association, Long Term Evolution (LTE)™, or more), WiFi (including 802.11 a/b/g/n/ac), WiMAX, Bluetooth, WiGig, and more.

[0023] The exemplary system 100 shown in FIG. 1 also includes the image processing system 104. As described in greater detail below, the image processing system 104 may determine a character count for an image received by the image processing system 104. The image processing system 104 may determine the character count for the image without recognizing the content or identity of characters in the image.

[0024] In some implementations, the image processing system 104 may include a communication interface 110, processing circuitry 112, and a user interface 114. The processing circuitry 112 of the image processing system 104 may perform any functionality associated with the image processing system 104, including any combination of the image processing techniques and methods described below. In one implementation, the processing circuitry 112 includes one or more processors 116 and a memory 120. The memory 120 may store image processing instructions 122 and character count parameters 124. The character count parameters 124 may include any parameters, settings, configurations, or criteria that control how the processing circuitry 112 determines process an image, including determining of a character count for the image. In some variations, the electronic device 102, such as a mobile device, may additionally or alternatively implement any of the functionality of the processing circuitry 112 described herein.

[0025] The processing circuitry 112 may process any digital image that may include text. As examples, the processing circuitry 112 may process an image of any type of financial document, including negotiable instruments such as personal checks, business checks, money orders, promissory notes, certificate of deposits, and more. As additional examples, the processing circuitry 112 may process images of any other type, such as a image of any type of insurance form or document, tax documents (e.g., form 1040), employment forms, savings bonds, traveler's checks, job applications, any type of bill, such as an automotive repair bill or medical bill, a remittance coupon, and images of many more types.

[0026] FIG. 2 shows an example of a financial document image 200 that the processing circuitry 112 may determine a character count for. A financial document image may refer to an image that includes any portion of a financial document. In particular, the example in FIG. 2 shows an image of a check, though the financial document image 200 may take various forms. The processing circuitry 112 may receive the financial document image 200 and apply any number of image processing functions prior to performing a character count process for the financial document image. For instance, the processing circuitry 112 may determine one or more corners or edges of the financial document image 200, deskew the image 200 to remove distortions, adjust the contrast of the image 200 to ease subsequent processing, apply any number of image cleaning algorithms, de-blur the image 200, binarize the image into black and white pixels, and more. As another example, the processing circuitry 112 may adjust the size of the image 200 to a particular height or width (e.g., in pixels) or to a particular proportion of the originally received image (e.g., shrink the image 200 to ¼ of the original size). The processing circuitry 112 may resize the image such that expected text in the financial document image (e.g., text of a particular font, particular font size, particular font size range, etc.) is of a particular pixel height, width, or within a particular pixel range.

[0027] The processing circuitry 112 may convert the financial document image 200 into a pixel array for processing. To illustrate, the financial document image 200 in FIG. 2 includes the image portion labeled as 210. The exemplary image portion 210 includes a number of pixels binarized into either white pixels or black pixels, as seen in the expanded image portion 210 reproduced and expanded in FIG. 2 above the financial document image 200. The processing circuitry 112 may represent the image portion 210 as a pixel array of black and white pixels. As one example, the processing circuitry 112 may represent the financial document image 200, including the image portion 210, as a two-dimensional array of array values, where an array value of ‘0’ indicates a white pixel and an array value of ‘1’ indicates a black pixel. The processing circuitry 112 may perform any of the image processing steps described above in preparation of determining the character count for the financial document image 200.

[0028] Exemplary processes through which the processing circuitry 112 may determine or estimate a character count for the financial document image 200 are presented next. First, the processing circuitry 112 may identify one or more text chunks in the financial document image 200, for example as described through FIGS. 3-8. Next, the processing circuitry 112 may determine a character count for some or all of the identified text chunks, for example as described through FIGS. 9-10.

Identifying Text Chunks

[0029] The processing circuitry 112 may identify one or more text chunks in the financial document image 200. A text chunk may refer to a particular portion of set of pixels of the financial document image 200 that may contain one or more text characters. In doing so, the processing circuitry 112 may evaluate the financial document image 200 to determine the likelihood particular portions (e.g., particular pixels) of the image 200 correspond to the boundary of a character, such as any edge, a top right corner, a top left corner, or other boundary portion of a text character.

[0030] In particular, the processing circuitry 112 may determine a likelihood that an image pixel corresponds to or is within a particular distance from the top left corner of a character. In some variations, the processing circuitry 112 may apply a top left scoring algorithm for pixels in the financial document image 200 to specify this likelihood. In scoring a particular pixel, the top left scoring algorithm may account for any number of other pixels surrounding the particular pixel. One such example is presented next in FIG. 3.

[0031] FIG. 3 shows an example of a top left scoring grid 301 for an image pixel 302. A scoring grid of pixels may represent a set of pixels surrounding a particular pixel, and the processing circuitry 112 may use the top left scoring grid 301 to apply a top left scoring algorithm to the image pixel 302. The top left scoring grid 301 may take various sizes or shapes, which may be specified through the character count parameters 124. The character count parameters 124 may, for instance, specify a height and/or width of the top left scoring grid 301. As shown in FIG. 3, the processing circuitry 112 determines the top left scoring grid 301 as a 6 by 6 array of pixels with the image pixel 302 being evaluated by the top left scoring algorithm positioned as the top left pixel of the top left scoring grid 301.

[0032] The character count parameters 124 may specify dimensions for a scoring grid according to any number of factors, some of which are presented next. The processing circuitry 112 may resize the financial document image 200 such that expected text of the image has a particular size, e.g., a MICR line, courtesy amount line, or other particular text in the financial document image 200 has particular pixel height, width, or pixel size range. The character count parameters 124 may specify, for example, dimensions for the top left scoring grid 301 such that the top left scoring grid 301 (or a non-padded portion thereof as discussed in greater detail below) covers a predetermined portion of an expected text character in the financial document image 200. As another variation, the character count parameters 124 may specify a scoring grid size that covers ⅓ the width an expected text character and ½ the height of an expected text character, which may be specified in pixels.

[0033] In some implementations, the character count parameters 124 specify the dimensions of a scoring grid to account for a particular pixel density of the financial document image 200. For instance, the character count parameters 124 may specify a particular dimension (e.g., 6 pixels wide by 9 pixels high) for scoring grid given a particular pixel density of the image 200 (e.g., for a 200 Dots-Per-Inch image). Additionally or alternatively, the character count parameters 124 may specify a scoring grid dimension to account for a minimum expected font size or minimum relevant font size in an image, for which the pixel size may vary depending on how the image was resized by the processing circuitry 112.

[0034] The processing circuitry 112 may determine a top left score for a pixel. The scoring algorithm may implement a scoring range indicative of the likelihood that the image pixel 302 corresponds to a top left corner of a character or is within a particular padded distance from the top left corner of a character. With regards to a padded distance, the top left scoring algorithm may include a padding parameter. The padding parameter may specify a particular padding of white pixels that surround the top left corner of a character. For example, with a padding parameter value of 2, the image pixel 302 may have an increased top left score when the image pixel 302 is two pixels above and two pixels to the left of a top left corner pixel of a text character. For top left scoring, the character count parameters 124 may specify a top padding parameter, a left padding parameter, or both.

[0035] The processing circuitry 112 may determine a top left score for the image pixel 302 according to the distribution of white and/or black pixels in the top left scoring grid 301. With padding, the scoring algorithm may indicate a higher likelihood of the image pixel 302 corresponding to a top left corner of a character when particular portions of the top left scoring grid 301 are white pixels, e.g. a white padded portion of the top left scoring grid 301. Along similar lines, the scoring algorithm may indicate a higher likelihood of a pixel corresponding to the top left corner when particular portions of the top left scoring grid 301 are black pixels, e.g., a black character portion. As one example, for a padding parameter value of 2 (for both top and left), the character count parameters 124 may specify an ideal distribution of white and black pixels in a 6×6 top left scoring grid 301 as the following configuration:

[0036]

	W	W	W	W	W	W
	W	W	W	W	W	W
	W	W	B	B	B	B
	W	W	B	B	B	B
	W	W	B	B	B	B
	W	W	B	B	B	B

The ideally white (W) pixels in the above ideal configuration may form the white padded portion for determining a top left score and the ideally black pixels (B) may form the black character portion for determining a top left score. When determining a top left score for a pixel, the processing circuitry 112 may determine the proportion of the white padded portion of the top left scoring grid 301 for that pixel that includes white pixels and the proportion of the black character portion that includes black pixels.

[0037] The processing circuitry 112 may apply weights when evaluating the pixels in the top left scoring grid 301. That is, the processing circuitry 112 may give more or less weight when a particular pixel in a particular position in the top left scoring grid 301 is either white or black. For instance, the character count parameters 124 may specify greater weight for pixels that are closer to a particular pixel, edge, or region in the top left scoring grid 301. One exemplary weighting for a 6×6 top left scoring grid 301 with a padding parameter value of 2 is as follows:

[0038]

	1	1	1	1	1	1
	1	2	2	2	2	2
	1	2	4	4	4	4
	1	2	4	3	3	3
	1	2	4	3	2	2
	1	2	4	3	2	1

In the above weighting, white pixels in the white padded portion are given a weight (e.g., multiplier) by 1 or 2. Black pixels in the white padded portion may be given a weight of 0. Black pixels in the black character portion are given a weight of 4, 3, 2 or 1, as shown by the underlined weights for pixels in the black character portion. White pixels in the black character portion may be given a weight of 0. In that regard, the processing circuitry 112 may determine a weighted proportion of white pixels in the white padded portion (e.g., a white padded score), for example by dividing a weighted sum for the pixels in a white padded portion by the ideal weighted value for the white padded portion. The processing circuitry 112 may determine a black character score in a consistent manner as well.

[0039] To illustrate, the processing circuitry 112 may determine the top left score for the image pixel 302 with the particular top left scoring grip 301 depicted in FIG. 3 according to a padding value of 2 and the exemplary weights shown above. The processing circuitry 112 may calculate a white padded portion score and black character score for the top left scoring grid 301. In particular, the white padded portion of the top left scoring grid 301 shown in FIG. 3 has pixel array of:

[0040]

	W	W	W	W	W	W
	W	W	W	W	W	W
	W	B	—	—	—	—
	W	B	—	—	—	—
	W	B	—	—	—	—
	W	B	—	—	—	—

Applying the weights for white and black pixels in the white padded portion, the processing circuitry 112 may determine the following weighted values for the white padded portion:

[0041]

	1	1	1	1	1	1
	1	2	2	2	2	2
	1	0	—	—	—	—
	1	0	—	—	—	—
	1	0	—	—	—	—
	1	0	—	—	—	—

Summing the weighted values, the processing circuitry 112 may determine the weighted sum for the white padded portions as 21. The processing circuitry 112 may identify the ideal weighted value (e.g., when all the pixels in the white padded portion are white) as 29. Accordingly, the processing circuitry 112 may determine the white padded score of the top left scoring grid 301 shown in FIG. 3 as 21/29=0.72.

[0042] The processing circuitry 112 may similarly determine a black character score of the top left scoring grid 301. The processing circuitry 112 may apply the exemplary weighting shown above for the black character portion of the top left scoring grid 301 in FIG. 3. In doing so, the processing circuitry 112 may determine the following weighted values for the black character portion:

[0043]

	—	—	—	—	—	—
	—	—	—	—	—	—
	—	—	4	0	0	0
	—	—	4	3	0	3
	—	—	4	3	2	2
	—	—	4	3	2	1

In this example, the processing circuitry 112 may determine the weighted sum for the black character portion as 35 and the ideal weighted sum (e.g., when all the pixels in the black character portion are black) as 50. Accordingly, the processing circuitry 112 may, in one implementation, determine the weighted black character score of the top left scoring grid 301 shown in FIG. 3 as 35/50=0.70.

[0044] The processing circuitry 112 may additionally apply weights when accounting for the white padded score and the black character score. When weighted equally, the processing circuitry 112 may determine the top left score of the image pixel 302 as the average the white padded score and black character score. In this example, the processing circuitry 112 determines the top left score of the image pixel 302 as (0.5)*0.72+(0.5)*0.70=0.71, as shown in FIG. 3. In other variations, the processing circuitry 112 may apply a greater weight to the white padded score than the black character score or vice versa.

[0045] The processing circuitry 112 may determine the top left score for the image pixel 302 as well as for any number of other pixels in the financial document image 200. The top left score determination method above may provide an quick and efficient method for determining the likelihood a particular pixel corresponds to the top left corner of a text character. The processing circuitry 112 may determine a respective top left score for pixels and identify pixels with a greater likelihood of corresponding to the top left corner of a character without, for example, performing edge detection processes or other processing-intensive processes.

[0046] While some particular examples have been presented above, the character count parameters 124 may specify any number of different configurations for determining a top left score, including varying height and width dimensions for the top left scoring grid 301, varying padding parameter values (including top padding, left padding, or both), as well as varying weight configurations, such as weights applied to particular pixels in the white padded portion or the black character portions, or to the white padded and black character scores. Another weighting configuration for a 6×9 top left scoring grid 301 with a top and left padding of 3 may be as follows (with black character portion weights underlined):

[0047]

	1	1	1	1	1	1
	1	2	2	2	2	2
	1	2	3	3	3	3
	1	2	3	3	2	1
	1	2	3	2	2	1
	1	2	3	1	1	1
	1	2	3	1	1	1
	1	2	3	1	1	1
	1	2	3	1	1	1

In this example, the white padded portion is weighted along the edge of the top and left edges of a character and the black character portion is weighted to emphasize the top left pixel of the character. The character count parameters 124 may implement any number of varying configurations through which the processing circuitry 112 determines the top left score for pixels in the financial document image 200.

[0048] FIG. 4 shows an example of a top left score array 400. In particular, the top left score array 400 shown in FIG. 4 includes top left scores for a row of pixels in the financial document image 200, e.g., after the processing circuitry 112 has determined the top left scores for that particular row of pixels. Along similar lines, the processing circuitry 112 may determine the top left scores for the remaining pixels of the financial document image 200 as well. In the exemplary top left score array 400 shown in FIG. 4, the processing circuitry 112 may determine the respective top left score for pixels in the financial document image 200 according to the following parameters: using a 6×9 pixel dimension for a top left scoring grid, a padding parameter value of 2, and an equal weight (e.g., half or 0.5) for the white padded score and black character score.

[0049] The processing circuitry 112 may determine a respective top left score for some or all of the pixels in the financial document image 200. For example, the processing circuitry 112 may abstain or forego determining the top left score for a pixel when the pixel is in a particular region of the financial document image 200, e.g., in the bottom-most row of the image 200, within a predetermined number of rows from the bottom-most row, in the right-most column of the image 200, or within a predetermined number of rows from the right-most column. As another example, the processing circuitry 112 may selectively determine the top left scores for pixels within a predetermined pixel distance from an interest region of the financial document image 200, such as the Magnetic Ink Character Recognition (MICR) location of a check, from a particular form field of an insurance document, and the like.

[0050] Before, during, or after determining the top left score for pixels in the financial document image 200, the processing circuitry 112 may determine a top right score for one or more pixels in the financial document image 200. In that regard, the processing circuitry 112 may determine a likelihood that an image pixel corresponds to or is within a particular pixel distance from the top right corner of a text character. The processing circuitry 112 may apply a scoring algorithm similar in many respects to top left scoring algorithm described above, but with any number of variances. For example, the configuration, weights, and other parameters specified by the character count parameters 124 for determining the top right score may be vertically mirrored from those used for determining a top left score.

[0051] The character count parameters 124 may specify distinct parameters through which the processing circuitry 112 determines a top right score for a pixel. In that regard, the character count parameters 124 may specify different configurations for a top right scoring grid as compared to the top left scoring grid, including differences in scoring grid dimensions, weights applied to pixels within a top right scoring grid, etc. In particular, the processing circuitry 112 may use a top padding parameter value and/or right padding parameter value in determining the top right score for a pixel, but not a left padding parameter value (as compared to the top left score determination parameters that may include a top left padding parameter value but not a top right padding parameter value).

[0052] FIG. 5 shows an example of a top right scoring grid 501 for an image pixel 502. The top right scoring grid 501 in FIG. 5 has a height and width of 6 pixels and a padding parameter value of 2. In this example, the character count parameters 124 may specify an ideal distribution of white and black pixels in the 6×6 top right scoring grid 501 as the following configuration:

[0053]

	W	W	W	W	W	W
	W	W	W	W	W	W
	B	B	B	B	W	W
	B	B	B	B	W	W
	B	B	B	B	W	W
	B	B	B	B	W	W

[0054] The ideally white (W) pixels in the above ideal configuration may form the white padded portion for determining the top right score for a particular pixel, e.g., the top right pixel of the top right scoring grid 501. The ideally black (B) pixels in the above ideal configuration may form the black character portion for determining the top right score. One exemplary weighting for a 6×6 top right scoring grid 501 with a padding value of 2 is as follows (with black character weights underlined):

[0055]

	1	1	1	1	1	1
	2	2	2	2	2	1
	4	4	4	4	2	1
	3	3	3	4	2	1
	2	2	3	4	2	1
	1	2	3	4	2	1

As seen, this exemplary weighting for a 6×6 top right scoring grid 501 is vertically mirrored from the exemplary weighting for a 6×6 top left scoring grid 301 discussed above.

[0056] The processing circuitry 112 may apply the weights to the top right scoring grid 501 for the image pixel 502 specifically shown in FIG. 5, which may be represented by the following pixel array (pixels in the black character portion underlined):

[0057]

	W	W	W	W	W	W
	W	W	W	W	W	W
	W	W	B	B	W	W
	B	W	B	B	W	W
	B	B	B	B	W	W
	B	B	B	B	W	W

The processing circuitry 112 may determine the weighted sum of the white padded portion to be 29 and the ideal weighted sum to be 29. In this example, the processing circuitry 112 determines the white padded score as 29/29=1.0. Following consistent respective calculations, the processing circuitry 112 may determine the black character score as 0.70. The processing circuitry 112 may, for example, apply the same weight to each score and determine the top right score of the image pixel 502 as (0.5)*1.0+(0.5)*(0.70)=0.85, as shown in FIG. 5. In a similar way, the processing circuitry 112 may determine a respective top right score for some or all of the pixels in the financial document image 200.

[0058] The processing circuitry 112 may determine top right scores and top left scores for some or all of the pixels in the financial document image 200. For a particular pixel, the processing circuitry 112 may determine a top right score for the particular pixel, a top left score for the particular pixel, or both. After determining top right and top left scores for pixels of the financial document image 200, the processing circuitry 112 may use the determined top right scores and top left scores to identify text chunks in the financial document image 200.

[0059] The processing circuitry 112 may identify a text chunk by determining one or more boundary pixels or edges for the text chunk. As one exemplary process described in greater detail below, the processing circuitry 112 may determine a top left pixel of the text chunk, a top right pixel of the text chunk, and a bottom edge of the text chunk. In that regard, the processing circuitry 112 may sequentially consider pixels in the financial document image 200 to identify a boundary of a text chunk. For example, the processing circuitry 112 may start the sequential processing of pixels for the text chunk determination process at the top left pixel of the financial document image 200. Or, the processing circuitry 112 may start with a pixel belongs to a particular portion of the financial document image 200, e.g., a MICR line portion of a check image.

[0060] The processing circuitry 112 may identify pixels or boundaries for a text chunk according to any number of chunk boundary criteria, which may be specified by the character count parameters 124. For a current pixel being considered for text chunk identification, the processing circuitry 112 may first determine whether the current pixel is already part of a previously determined text chunk. If so, the processing circuitry 112 may exclude the current pixel from belonging to another text chunk and proceed to a subsequent pixel for consideration.

[0061] When a current pixel is not part of a previously determined text chunk, the processing circuitry 112 may determine whether the current pixel meets chunk boundary criteria for a top left pixel of the text chunk. The processing circuitry 112 may identify the current pixel as the top left pixel for a text chunk when the top left score of the current pixel is equal to or exceeds a top left score threshold, such as a top left score threshold of 0.65 in some implementations. Accordingly, the processing circuitry 112 may identify a top left pixel for the text chunk without performing additional or more complicated image processing techniques, e.g., without performing edge detection algorithms. After determining a top left pixel for a text chunk, the processing circuitry 112 may determine the top right pixel for the text chunk.

[0062] The processing circuitry 112 may determine a top right pixel for the text chunk by evaluating pixels to the right of the determined top left pixel of the text chunk. In that regard, the processing circuitry 112 may determine a set of potential top right pixels based on the top left scores of the pixels being evaluated, e.g., top right candidate pixels. In one implementation, the processing circuitry 112 determines the top right candidate pixels for the text chunk as a set of consecutive pixels with a top left score below a top left score threshold, as set by the character count parameters 124. The character count parameters 124 may specify the same or a different top left score threshold used for identifying the top left pixel of the text chunk and the top right candidate pixels of the text chunk.

[0063] To illustrate, the processing circuitry 112 may start at the determined top left pixel of the text chunk and sequentially consider pixels to the right of the top left pixel. When the current pixel has a top left score less than the top left score threshold for identifying a top right pixel (e.g., 0.65), the processing circuitry 112 may increment a counter value and continue to the next pixel. When the current pixel has a top left score equal to or greater than the top left score threshold for identifying a top right pixel (e.g., 0.65), the processing circuitry 112 may reset the counter to 0 and continue to the next pixel. When the counter value reaches a counter threshold value (e.g., 13), the processing circuitry 112 may identify a number of previously considered pixels equal to the counter threshold value as the top right candidate pixels (e.g., the 13 previously considered pixels when the counter threshold value is 13). An exemplary iteration of this process is presented in FIG. 6.

[0064] FIG. 6 shows an exemplary flow 600 for determining a top right pixel of a text chunk. In the exemplary flow 600 in FIG. 6, the processing circuitry 112 reads the character count parameters 124, which may specify a top left score threshold for identifying the top right pixel as 0.65 and the counter threshold value as 13. As the processing circuitry 112 evaluates successive pixels, the processing circuitry 112 either increments a counter value when the top left score of the pixel is less than 0.65 or resets the counter value when the top left score of the pixels is equal to or greater than 0.65. In this example, when the processing circuitry 112 identifies thirteen (13) consecutive pixels with a top left score less than 0.65, the processing circuitry 112 determines the top right candidate pixels 610 for the text chunk.

[0065] The processing circuitry 112 may determine the top right pixel for the text chunk from among the top right candidate pixels 610. In some implementations, the processing circuitry 112 identifies the pixel with the highest top right score from among the top right candidate pixels 610 as the top right pixel for the text chunk. In the example shown in FIG. 6, the processing circuitry 112 determines the pixel with the top right score of 0.83 as the top right pixel 620 for the text chunk. The processing circuitry 112 may identify a top right pixel for the text chunk without performing additional or more complicated image processing techniques, e.g., without performing edge detection algorithms.

[0066] After determining a top left and top right corner for a text chunk, the processing circuitry 112 may determine a bottom edge of the text chunk. In doing so, the processing circuitry 112 may consider rows of pixels below the particular row of pixels formed by and between the top left and top right pixels, e.g., the top row of the text chunk. The processing circuitry 112 may identify a first row of pixels below the top row of the text chunk formed by the top left and top right pixels with a proportion of white pixels that exceeds a bottom edge threshold. In some variations, the character count parameters 124 may set the bottom edge threshold at 90%, for example. The processing circuitry 112 may determine the bottom edge of the text chunk as the upper edge of the identified first row of pixels with a proportion of white pixels that exceeds the bottom edge threshold.

[0067] In determining the bottom edge of the pixel chunk, the processing circuitry 112 may ignore or not consider a number of pixel rows at the top of the text chunk equal to the padding parameter value. For example, when the padding parameter value is set to 2, the processing circuitry 112 may not consider the top two rows of pixels formed between the top left and top right pixels when identifying the bottom edge of the text chunk. Put another way, the processing circuitry 112 may forego considering the top row of pixels that includes the top left and top right pixels and the next row of pixels directly below the top row when the padding parameter value is 2. Accordingly, the processing circuitry 112 may determine a text chunk formed by a top left pixel, a top right pixel, and a bottom edge. The processing circuitry 112 may further process the text chunk as well, some examples of which are shown in FIG. 7.

[0068] FIG. 7 shows examples of text chunks 710, 720, and 730 the processing circuitry 112 may determine. The processing circuitry 112 may determine the text chunk labeled as 710 with boundaries set by the top left pixel 711, top right pixel 712, and the bottom edge 713 using any of the methods or techniques described above. In particular, in determining the text chunk 710, the character count parameters 124 may specify a padding parameter value of 2, and accordingly, the processing circuitry 112 may forego considering the first two rows of pixels in the text chunk 710 when determining the bottom edge 713. As seen in FIG. 7, the row of pixels below the bottom edge 713 is composed entirely of white pixels, and the processing circuitry 112 may identify this row as the first row of pixels that exceed a bottom edge threshold. Accordingly, the processing circuitry 112 may determine the bottom edge 713 of the text chunk 710 as the upper edge of this row of white pixels exceeding the bottom edge threshold.

[0069] In some variations, the processing circuitry 112 may pad the bottom edge 713 of a determined text chunk 710. In that regard, the processing circuitry 112 may add a number of rows of white pixels below the bottom edge 713 as set by the character count parameters 724. In the example shown in FIG. 7, the processing circuitry 112 determine the text chunk 720 by padding the text chunk 710 with two rows of white pixels, e.g., through setting the padded bottom edge 723 as the bottom edge of a text chunk.

[0070] In some variations, the processing circuitry 112 may adjust the left or right edges of a determined text chunk. For example, the processing circuitry 112 may pad the left edge of a text chunk, right edge of a text chunk, or both as similarly described above with regards to padding the bottom edge 713 of the text chunk 710. The processing circuitry 112 may additionally or alternatively adjust the left or right edges of a text chunk to include a chunk extension. In FIG. 7, the processing circuitry 112 may determine the text chunk 730 by adjusting the text chunk 720 to include the chunk extension 740.

[0071] The processing circuitry 112 may determine a chunk extension for a text chunk. In doing so, the processing circuitry 112 may consider the columns of pixels to the right or left of an edge of a text chunk and determine occurrence of a threshold number of consecutive white pixel columns, e.g., a consecutive number of columns each or which and/or collectively have a proportion of white pixels that exceed an extension threshold, such as 90% white pixels. The columns considered by the processing circuitry 112 may be to the left or right of the text chunk and have the same height as the text chunk. The processing circuitry 112 may determine the chunk extension as the section of pixels between the edge of the text chunk and the identified consecutive white pixels columns.

[0072] As one particular example shown in FIG. 7, the processing circuitry 112 may determine the text chunk 720. Then, the processing circuitry 112 may determine the chunk extension 740 when, for example, the next threshold number of (e.g., the next 20) pixel columns to the right of the chunk extension 740 in the financial document image 200 are each 90% (or more) white. Accordingly, the processing circuitry 112 may determine the text chunk 730 by appending the chunk extension 740 to the text chunk 720. As such, the processing circuitry 112 may identify the top right pixel 732 in the chunk extension 740 as part of the boundary of the text chunk 730.

[0073] In processing the financial document image 200, the processing circuitry 112 may determine the text chunks 710 and 720 shown in FIG. 7 as intermediate text chunks and the text chunk 730 as the determined text chunk. However, the character count parameters 124 may include any number of additional or alternative parameters for determining intermediate text chunks before obtaining a determined text chunk.

[0074] FIG. 8 shows an example of logic 800 for identifying one or more text chunks in an image. The logic 800 may be implemented in hardware, software, or both. For example, the processing circuitry 112 may implement the logic 800 in software as the image processing instructions 122.

[0075] The processing circuitry 112 may read the character count parameters 124 (802) and receive an image (804). The image may, for example, be a financial document image 200 such as a check or insurance form. The processing circuitry 112 may perform various pre-processing on the image, such as cleaning up the image, resizing the image to control the text size of expected text in the image (e.g., text of a MICR line or courtesy line in a check), binarizing the image, or converting the image to a pixel array.

[0076] The processing circuitry 112 may determine a respective top left score for some or all of the pixels in the image (804). The processing circuitry 112 may determine a respective top right score for some or all of the pixels in the image (806), including for pixels different from the pixels for which the processing circuitry 112 determines respective top left scores. In some implementations, the processing circuitry 112 may determine the top left scores and/or top right scores for specific portions (e.g., interest regions) of the image, and decline or skip determining the top left and/or top right scores for pixels outside of the interest regions of the image. As one example, the character count parameters 124 may specify a MICR line region, upper left hand corner, amount region on the middle right side, or other portions of a check image as interest regions.

[0077] Upon determining the top left and top right scores for image pixels in the image, the processing circuitry 112 may sequentially process pixels in the image. In that regard, the processing circuitry may determine whether any additional pixels in the image remain for processing (810), e.g., whether any unprocessed pixels remain in the interest region(s) of the image. If so, the processing circuitry 112 may set a current pixel (812). The processing circuitry 112 may set the current pixel as the next pixel in a pixel processing ordering. For example, the processing circuitry 112 may process pixels ordering from left to right and row by row from the top left corner of the image to the bottom right corner of the image or interest region.

[0078] For a current pixel, the processing circuitry 112 determines whether the current pixel is already part of a previously formed text chunk (816). For example, the processing circuitry 112 may access a listing of determined text chunks for the image, and determine whether the pixel is already part of another determined text chunk. If so, the processing circuitry 112 may proceed to consider a subsequent pixel of the image or interest region, if any remain (810).

[0079] The processing circuitry 112 may determine the boundaries of a text chunk. In that regard, the processing circuitry 112 may determine a top left corner for the text chunk. The processing circuitry 112 may determine, for example, whether the top left score of the current pixel exceeds (or alternatively, is equal to or greater than) a top left score threshold, which may be set by the character count parameters 124. When the top left score of the current pixel does not exceed the top left score threshold, the processing circuitry 112 may proceed to consider a subsequent pixel of the image or interest region, if any remain (810). When the top left score of the current pixel exceeds the top left score threshold, the processing circuitry 112 may set the current pixel as the top left corner of the text chunk (820).

[0080] Continuing the boundary determination for a text chunk, the processing circuitry 112 may determine a top right pixel for the text chunk (822) through any of the methods or techniques described above. For example, the processing circuitry 112 may determine a set of top right candidate pixels from the image, and identify the top right corner of the text chunk as the pixel from among the top right candidate pixels with the highest top right score (e.g., the pixel most likely to correspond to the top right corner or a text character as specified by top right score). The processing circuitry 112 may also determine the bottom edge of the text chunk (824) through any of the processes and techniques described above.

[0081] The processing circuitry 112 may further adjust the boundaries of a text chunk. In some variations, the processing circuitry 112 determines one or more chunk extensions (826) through which to extend the left edge or right edge (or both) of a text chunk. The processing circuitry 112 may additionally or alternatively pad the text chunk with white pixels, for example as specified by padding parameter(s) in the character count parameters 124. Using any combination of the techniques, process, or steps described above, the processing circuitry 112 may determine a text chunk.

[0082] Upon determining a text chunk, the processing circuitry 112 may validate the text chunk (830). As an exemplary validation, the processing circuitry 112 may determine whether the height of text chunk (e.g., pixel height) exceeds a minimum height threshold (e.g., 10 pixels). As another example, the processing circuitry may determine whether the height of the text chunk is within a maximum height threshold (e.g., 50 pixels). In some variations, the processing circuitry 112 may validate that all (or a threshold percentage) of the pixels in the text chunk are not a part of another determined text chunk. In these variations, the processing circuitry 112 may access a listing of previously determined text chunks to determine whether pixels of the text chunk belong to any of the previously determined text chunks. When the text chunk passes the validation process, the processing circuitry 112 may store the text chunk (832), e.g., by storing an indication of the text chunk in the determined text chunk listing. The indication may, for example, take the form of a database or data structure entry and may specify the boundary and/or pixels belonging to the associated text chunk. Then, the processing circuitry 112 may consider the subsequent pixel of the image or interest region, if any remain (810). When the text chunk fails the validation, the processing circuitry 112 may discard the text chunk and not store the text chunk in the determined text chunk listing. That is, the processing circuitry 112 may proceed to consider the subsequent pixel of the image or interest region (810) without storing an indication of the text chunk.

Determining Character Count

[0083] After identifying text chunks in an image, e.g., a financial document image 200, the processing circuitry 112 may determine a character count for the text chunks. As described in greater detail below, the processing circuitry 112 may determine the character count for a text chunk without specifically recognizing the identity or content of any particular text characters in the text chunk. For example, the processing circuitry 112 may determine the character count for the text chunk without performing any character recognition techniques, such as Optical Character Recognition (OCR) or other similar character recognition techniques.

[0084] FIG. 9 shows an exemplary character count 900 in a text chunk 910. The processing circuitry 112 may perform the character count 900 to determine a character count for the text chunk 910 without recognizing any particular characters in the text chunk 910. That is, the text chunk 910 shown in FIG. 9 includes the text “Wilmington, Del.” and the processing circuitry 112 may obtain a character count for the text chunk 910 without recognizing the letters or the comma within the text chunk 910.

[0085] The processing circuitry 112 may process the text chunk 710 to determine a character start column and a corresponding character end column. To do so, the processing circuitry 112 may start at the leftmost pixel column of the text chunk and sequentially consider pixel columns in text chunk 710. The processing circuitry 112 may identify a character start column when the number of black pixels in a current column exceeds a black column threshold, which may be specified in the character count parameters 124 as a number of pixels or percentage, for example. In the specific example shown in FIG. 9, the processing circuitry 112 identifies the character start column when a current pixel column has at least one black pixel or greater than 0% of black pixels. The processing circuitry 112 may identify the character start columns shown in FIG. 9 that are marked with the dotted arrows and labeled with a corresponding character start number.

[0086] Upon identifying a character start column, the processing circuitry 112 may continue to sequentially consider pixel columns to the left of the character start column to identify a corresponding character end column. The processing circuitry 112 may identify the corresponding character end column as the first pixel column to the right of the character start column with white pixels that exceed a white column threshold. The processing circuitry 112 may identify a character end column when a current pixel column has less than 2 black pixels, for example. As seen in the exemplary text chunk 910 in FIG. 9, the processing circuitry 112 identifies the character end columns marked with the non-dotted arrows and labeled with a corresponding character end number. After identifying a corresponding character end column for a particular character start column, the processing circuitry 112 may increment a character count value and continue sequentially considering pixel columns of the text chunk 910 to determine a next character start column. The processing circuitry 112 may continue the character count process until reaching the end of the text chunk 910, e.g., considering the rightmost column of the text chunk 910. In the example shown in FIG. 9, the processing circuitry 112 determines character count of the text chunk 910 be 13, and determines the character count without recognizing the content or identity of any of the characters in the text chunk 910.

[0087] FIG. 10 shows an example of logic 1000 for obtaining a character count for a text chunk. The logic 1000 may be implemented in hardware, software, or both. For example, the processing circuitry 112 may implement the logic 1000 in software as the image processing instructions 122.

[0088] The processing circuitry 112 may read the character count parameters 124 (1002) and obtain a text chunk (1004). In some implementations, the processing circuitry 112 may obtain the chunk by accessing a text chunk listing or data structure, which may provide, for example, an indication of the boundaries of a particular text chunk in an image. The text chunk may be in the form of a pixel array.

[0089] In determining a character count for a text chunk, the processing circuitry 112 may process one or more pixel columns in the text chunk. The processing circuitry 112 may determine a character start column in the text chunk and then a corresponding character end column. To do so, the processing circuitry 112 may process the pixel columns in the chunk in according to a pixel column processing ordering. For example, the processing circuitry 112 may process pixel columns in the text chunk in a sequential order from the left most pixel column to the right most pixel column. Accordingly, the processing circuitry 112 may determine whether any additional pixel columns in the text chunk remain for processing (1006). If so, the processing circuitry 112 may set the next pixel column in the pixel column processing ordering as the current column for determining a character start column (1008).

[0090] The processing circuitry 112 may identify a character start column when a current pixel column meets any number of character start column criteria. The processing circuitry 112 may identify a character start column based on a black column threshold, which may specify a percentage, proportion, or number of black pixels in a pixel column. Accordingly, the processing circuitry 112 may identify a character start column by determining whether the number or proportion of black pixels in the current pixel column exceeds a black column threshold (1010). If not, the processing circuitry 112 may consider the next pixel column in the text chunk for determining a character start column, if any remain (1006). When the number or proportion of black pixels in the current pixel column exceeds the black column threshold, the processing circuitry 112 identifies this particular pixel column as a character start column (1012).

[0091] The processing circuitry 112 may determine a corresponding character end column for the identified character start column. After identifying the character start column, the processing circuitry 112 may consider the next pixel column in the text chunk, if any remain (1014). If so, the processing circuitry 112 may set the next pixel column as the current column for determining a character end column (1016). The processing circuitry 112 may identify a character end column when a current pixel column meets any number of character end column criteria. In particular, the processing circuitry 112 may, for example, identify a character end column by determining whether the number or proportion of white pixels in the current pixel column exceeds a white column threshold (1018). If not, the processing circuitry 112 may consider the next pixel column in the text chunk for determining a character end column, if any remain (1014).

[0092] When the number or proportion of white pixels in the current pixel column exceeds the white column threshold, the processing circuitry 112 identifies this particular pixel column as a corresponding character end column to the previously determined character start column (1020). The processing circuitry 112 may increment a counter indicating the character count for the text chunk.

[0093] The processing circuitry 112 may continue processing the text chunk to determine character start columns and corresponding character end columns until no additional pixel columns remain (1006 or 1014). Then, the processing circuitry 112 may obtain the character count for the text chunk by reading the counter indicating the character count for the text chunk (1022).

[0094] FIG. 11 shows an example of logic 1100 that may be implemented in hardware, software, or both. For example, the processing circuitry 112 may implement the logic 1100 in software as the image processing instructions 122.

[0095] The processing circuitry 112 may read the character count parameters 124 (1102) and receive a financial document image 200 (1104). In some implementations, the character count parameters 124 may specify a character count threshold for the financial document image 200. The character count threshold may specify a minimum or maximum threshold number of characters in a financial document image 200 to meet particular quality criteria for processing the financial document image 200. Additionally, the character count parameters 124 may specify a particular character count threshold for different types of images, such as specific character count thresholds for business checks, personal checks, financial forms, remittance coupons, etc. As illustrative examples, the character count threshold for a business check may be set to 50 characters for personal checks and 100 characters for business checks. The character count parameters 124 may additionally or alternatively specify a particular character count threshold for particular regions (e.g., interest regions) of the financial document image 200 or any other image type the processing circuitry 112 may process.

[0096] The processing circuitry 112 may optionally perform image pre-processing techniques on the financial document image 200 (1106), including any of the pre-processing techniques described above. The processing circuitry 112 may determine a character count for the financial document image 200, for example by identifying one or more text chunks in the financial document image 200 (1008) and determining a character count for one or more of the identified text chunks (1110). To do so, the processing circuitry 112 may utilize any combination of the methods, flows, and techniques described above. The processing circuitry 112 may determine a character count for the financial document image 200 by summing the determined character count of text chunks in the financial document image 200.

[0097] The processing circuitry 112 may determine whether the character count for the financial document image 200 meets the character count criteria (1112). When the character count for the financial document image 200 fails the character count criteria, the processing circuitry 112 may instruct recapture of the financial document image (1114). For example, the processing circuitry 112 may send an image rejection message to an electronic device 102 used to capture the financial document image 200. The image rejection message may further instruct a user to recapture the image of financial document.

[0098] When the character count for the financial document image 200 meets the character count criteria, the processing circuitry 112 may perform further image processing. For example, the processing circuitry 112 may perform character recognition (e.g., OCR) on the financial document image 200 to recognize the characters on the financial document image 200. The processing circuitry 112 may perform further processing after character recognition, such as initiating a deposit process of a check represented by the financial document image 200, processing of a medical bill or financial form, etc.

[0099] The character count criteria may serve as an initial quality screen for incoming images received by the processing circuitry 112. By determining the character count of an image prior to performing subsequent image processing, the processing circuitry 112 may determine that the image is not overly cropped, and thus containing a character count less than a minimum threshold. Similarly, the character count criteria may be configured to prevent processing of overly blurry images, e.g., blurry images such that the processing circuitry 112 cannot determine enough character start and end columns and resulting in a character count less than a minimum threshold.

[0100] As discussed above, the character count parameters 124 may specify particular character count thresholds for interest regions of an image. Accordingly, the processing circuitry 112 may specifically identify text chunks and determine character counts for these interest regions instead of for the entire image. The processing circuitry 112 may determine the image passes the character count criteria when some or all of the particular character count criteria for the determined interest regions are met. As one example, the processing circuitry 112 may identify a MICR line portion of a check image as an interest region, and apply particular character count criteria for the MICR line portion, e.g., a minimum character count threshold. Additional exemplary interest regions may include high priority fields of a document, such as a social security number field, name field, address field, courtesy amount field, or any high priority region of an image received by the processing circuitry.

[0101] By performing combinations of the methods and techniques described above, the processing circuitry 112 may identify the character count for a financial document image 200 or other image without recognizing any particular character in the financial document image 200.

[0102] Although the example of a financial document 200 such as a check is provided by way of example above, the techniques discussed for identifying the presence, but not the specific identity or literal meaning, of characters or words, may be applied to any type of document. Other documents that may be analyzed with the techniques described herein include receipts, insurance documents, coupons, and so on. Specific portions of these documents may be targeted, or only characters of a particular font size may be included, for a given type of document. An advantage of the techniques discussed above is that the processing power and time for recognizing the presence, but not the specific identity, of characters or words may be less than that needed for actually identifying the individual letter, number or symbol. In other words, the knowledge that the captured image has chunks of text with a likelihood of four characters may be used rather than identifying those four characters as “abc3’ can provide a helpful filter for a system to determine if an expected type of document is being looked at. In this way, a system may quickly, and with less processing power, filter out unacceptable (e.g., overly cropped or blurry) documents.

[0103] In some implementations, an image processing system 104 may implement the processing circuitry 112 for performing any of the methods and techniques described above, including determining a character count for a financial document image 200 without recognizing any particular character in the financial document image 200. In other implementations, an electronic device 102, such as a mobile device, may implement the processing circuitry 112. In yet other implementations, the functionality of the processing circuitry 112 may be implemented, e.g., distributed, through a combination of the image processing system 104 and electronic device 102.

[0104] The methods, devices, and logic described above may be implemented in many different ways in many different combinations of hardware, software or both hardware and software. For example, all or parts of the system may include circuitry in a controller, a microprocessor, or an application specific integrated circuit (ASIC), or may be implemented with discrete logic or components, or a combination of other types of analog or digital circuitry, combined on a single integrated circuit or distributed among multiple integrated circuits. All or part of the logic described above may be implemented as instructions for execution by a processor, controller, or other processing device and may be stored in a tangible or non-transitory machine-readable or computer-readable medium such as flash memory, random access memory (RAM) or read only memory (ROM), erasable programmable read only memory (EPROM) or other machine-readable medium such as a compact disc read only memory (CDROM), or magnetic or optical disk. Thus, a product, such as a computer program product, may include a storage medium and computer readable instructions stored on the medium, which when executed in an endpoint, computer system, or other device, cause the device to perform operations according to any of the description above.

[0105] The processing capability described above may be distributed among multiple system components, such as among multiple processors and memories, optionally including multiple distributed processing systems. Parameters, databases, and other data structures may be separately stored and managed, may be incorporated into a single memory or database, may be logically and physically organized in many different ways, and may implemented in many ways, including data structures such as linked lists, hash tables, or implicit storage mechanisms. Programs may be parts (e.g., subroutines) of a single program, separate programs, distributed across several memories and processors, or implemented in many different ways, such as in a library, such as a shared library (e.g., a dynamic link library (DLL)). The DLL, for example, may store code that performs any of the system processing described above. While various embodiments of the systems and methods have been described, it will be apparent to those of ordinary skill in the art that many more embodiments and implementations are possible within the scope of the systems and methods. Accordingly, the systems and methods are not to be restricted except in light of the attached claims and their equivalents.