Systems and methods for motion adaptive filtering

[0001] The present application is related to commonly owned U.S. patent applications: (1) U.S. patent application Ser. No.______, entitled SYSTEMS AND METHODS FOR TEMPORAL SUBPIXEL RENDERING OF IMAGE DATA, and (2) commonly owned U.S. patent application Ser. No.______, entitled SUB-PIXEL RENDERING SYSTEM AND METHOD FOR IMPROVED DISPLAY VIEWING ANGLES, which are hereby incorporated herein by reference.

[0002] In commonly owned U.S. patent applications: (1) U.S. patent application Ser. No. 09/916,232 (the '232 application), entitled ARRANGEMENT OF COLOR PIXELS FOR FULL COLOR IMAGING DEVICES WITH SIMPLIFIED ADDRESSING, filed Jul. 25, 2001; (2) U.S. patent application Ser. No. 10/278,353 (the '353 application), entitled IMPROVEMENTS TO COLOR FLAT PANEL DISPLAY SUB-PIXEL ARRANGEMENTS AND LAYOUTS FOR SUB-PIXEL RENDERING WITH INCREASED MODULATION TRANSFER FUNCTION RESPONSE, filed Oct. 22, 2002; (3) U.S. patent application Ser. No. 10/278,352 (the '352 application) entitled IMPROVEMENTS TO COLOR FLAT PANEL DISPLAY SUB-PIXEL ARRANGEMENTS AND LAYOUTS FOR SUB-PIXEL RENDERING WITH SPLIT BLUE SUBPIXELS, filed Oct. 22, 2002; (4) U.S. patent application Ser. No. 10/243,094 (the '094 application), entitled IMPROVED FOUR COLOR ARRANGEMENTS AND EMITTERS FOR SUBPIXEL RENDERING, filed Sep. 13, 2002; (5) U.S. patent application Ser. No. 10/278,328 (the '328 application), entitled IMPROVEMENTS TO COLOR FLAT PANEL DISPLAY SUB-PIXEL ARRANGEMENTS AND LAYOUTS WITH REDUCED BLUE LUMINANCE WELL VISIBILITY, filed Oct. 22, 2002; (6) U.S. patent application Ser. No. 10/278,393 (the '393 application) entitled COLOR DISPLAY HAVING HORIZONTAL SUB-PIXEL ARRANGEMENTS AND LAYOUTS filed Oct. 22, 2002; (7) U.S. patent application Ser. No.______(the '______ application) entitled IMPROVED SUB-PIXEL ARRANGEMENTS FOR STRIPED DISPLAYS AND METHODS AND SYSTEMS FOR SUB-PIXEL RENDERING SAME, filed Jan. 16, 2003, novel subpixel arrangements are therein disclosed for improving the cost/performance curves for image display devices and herein incorporated by reference.

[0003] These improvements are particularly pronounced when coupled with subpixel rendering (SPR) systems and methods further disclosed in those applications and in commonly owned U.S. patent applications: (1) U.S. patent application Ser. No. 10/051,612 (the '612 application), entitled CONVERSION OF RGB PIXEL FORMAT DATA TO PENTILE MATRIX SUB-PIXEL DATA FORMAT, filed Jan. 16, 2002; (2) U.S. patent application Ser. No. 10/150,355 (the '355 application), entitled METHODS AND SYSTEMS FOR SUB-PIXEL RENDERING WITH GAMMA ADJUSTMENT, filed May 17, 2002; (3) U.S. patent application Ser. No. 10/215,843 (the '843 application), entitled METHODS AND SYSTEMS FOR SUB-PIXEL RENDERING WITH ADAPTIVE FILTERING, filed Aug. 8, 2002all patent applications and other references mentioned in this specification are herein incorporated by reference.

[0004] The accompanying drawings, which are incorporated in, and constitute a part of this specification illustrate exemplary implementations and embodiments of the invention and, together with the description, serve to explain principles of the invention.

[0005] FIGS. 1A and 1B depict a display screen with a word processing window open having text on the rendered in the window and scrolling down the screen.

[0006] FIG. 2 shows the performance curves of a liquid crystal display at the 100% illumination curve and the 50% illumination curve versus time.

[0007] FIG. 3 depicts one possible embodiment of a system made in accordance with the principles of the present invention.

[0008] FIGS. 4 and 5 are a flowchart of one embodiment of the techniques in accordance with the principles of the present invention.

[0009] Reference will now be made in detail to implementations and embodiments, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

[0010] As was described in the two related patent applications noted above, some panel technologiesmost notably liquid crystal displays (LCDs)exhibit color error on subpixel rendered text or other areas of high spatial frequency (HSF) when viewed by an observer at an off-nonmal axis viewing angle. Those related applications disclose systems and methods for correcting such color error from off-normal viewing angles.

[0011] Color errors from other than off-normal viewing angle may be noticeable by viewers on some LCDseven when observing from the normal axis to the display panel. For example, moving subpixel rendered text (or other areas of high spatial frequency) may produce color error while in motion. One example of this effect is scrolling text in a word processor application window. Depending on the panel technology (e.g. twisted nematic TN-LCD), the color error may be quite noticeableand possibly distracting to a userwhile scrolling the text. Of course, once the scrolling or motion stops, the color error typically ceases as the response time of TN LCD have time to catch up to the now-stationary text.

[0012] FIGS. 1A and 1B depict the situation in the one example noted. Display panel 100 is shown having a word processing application sending image data to the panel in a window 101. In the word processing window, there is some text 102 which is a paradigm example of high spatial frequency data. One point of the text 104 (e.g. an edge point of the character T) is at coordinate point (X,Y) on the panel in FIG. 1A. As the text is scrolled down, the edge point moves to an new coordinate point 106 (X, Y) on the panel in FIG. 1B. During the time the edge point was in transit between points 104 and 106, the T was visible on screen and was moving frame-by-frame down to its new point. If the response time of the panel's rendering technology (e.g. liquid crystal) is not sufficiently fast, then, if the T was being subpixel rendered at the time, color error may be noticeable.

[0013] FIG. 2 shows response curves for liquid crystal in going from either 100% or 50% illumination down to 0% illuminations (i.e. curves 202 and 204 respectively) and from 50% or 0% illumination up to 100% illumination (i.e. curves 206 and 208 respectively) and gives a better explanation as the nature of the problem. For example, when black text is rendered on display having a repeat subpixel grouping such as found in FIG. 5 of the related application, entitled SYSTEMS AND METHODS FOR TEMPORAL SUBPIXEL RENDERING OF IMAGE DATA,, the green subpixels switch from 100% to 0% while the red and blue pixels switch from 100% to 50%. During motion of black text, the green pixels are therefore switching from 100% to 0% to 100%while the red and the blue are switching from 100% to 50% to 100%. As is illustrated, the response time of the 100% to 0% is much faster than the 100% to 50%.

[0014] During motion of black text then, there will be an unbalanced condition of the brightness of red, green and blue pixels, which leads to color error. In fact, there will tend to be too much red and blue brightness which causes a magenta hue to the text. The transition from 0% to 100% is approximately the same as 50% to 100% so doesn't materially add to color error in this example. However, in other LCD modes, this transition could also have larger differences and will lead to color error during motion.

[0015] One embodiment to reduce the amount of color error on moving subpixel rendered text and other high spatial frequency image data is to employ an adaptive filter technique. An adaptive filter test may be used to detect motion of high-spatial-frequency edges in an image. When the moving edges are detected, subpixel rendering (SPR) of the text can be changed to a new state. After the moving edges are stationary, the SPR is turned back to the regular mode. Techniques such as those disclosed in the '843 application can be used to detect the edges and to detect the high frequency transitions in the data. A simple counter can be used with the SPR algorithm that counts the number of times an edge is detected in an image. Statistically, a large number of edges means that text is detected. If a low number of edges are detected, then the image is probably pictorial. Since this problem occurs primarily on edges of text, one embodiment might be to employ the change in filters for text only.

[0016] FIG. 3 depicts one system embodiment 300 for motion adaptive filtering. System 300 comprises a graphics subsystem 302 having a SPR subsystem 304 which comprises systems and methods of subpixel rendering source (such as those disclosed in the '612 application, '355 application, and the '843 application) image data 312 which is input into the graphics subsystem. The source image data may typically be generated by the operating system 314 or an application 316 and sent to the graphics subsystem for rendering on the display.

[0017] A memory 306 is available to SPR subsystem 304 to retain information about the number of and/or locations of points of high spatial frequency in the source image data. A timing controller (TCON) 308 is optionally provided to give timing commands to a display panel 310 which could be a LCD or any other technology having a suitably different response times vs. grey level to produce the color error discussed above. It will be appreciated that the system 300 is merely one possible embodiment to implement the techniques disclosed herein. For example, the SPR subsystem could be an application system integrated circuit (ASIC), field programmable gate array (FPGA), implemented entirely in software under control of a general processor control, or even implemented on the glass of the panel itself (particularly for low temperature polysilicon (LTPS) panels). Additionally, memory 106 could be implemented in RAM of any known species or any other known or future embodiment of memory. One embodiment comprises a graphical subsystem further comprising: a subpixel rendering subsystem; a memory coupled to said subpixel rendering subsystem for storing input image data in a plurality of image frames, and a processing subsystem that tests for moving text or other points of high spatial frequency and if the test indicated moving text or said other areas of high spatial frequency, sending signals to said subpixel rendering subsystem to change the subpixel rendering in successive frames of image data. It will be appreciated that the processing subsystem may be implemented integral or as a part of the subpixel rendering subsystem itself.

[0018] FIG. 4 shows one embodiment of implementing a technique for correcting such color errors. In essence, the embodiment involves a display system comprising a graphics subsystem having a subpixel rendering system. The system notes points of high spatial frequency in a first frame of image data; compares the points of high spatial frequency with corresponding points in a second frame of image data; and if a number of points have changed from high spatial frequency to low spatial frequency in the second frame meeting a certain threshold, changing the subpixel rendering on input image data.

[0019] The technique starts at step 402 where a image data point at coordinate (X,Y) is input into the SPR subsystem. The point is tested at step 404 to see if it is the point at the end of a frame. If yes, then the technique starts processing at step 406. If not, then the point is tested (via an adaptive filter or by any other means known now or in the future) whether the point is at the edge of a high spatial frequency area (e.g. text) at step 408. If it is not, then at step 410 the image data is incremented at the next coordinate point and returns to step 402. Of course, other SPR functions could be applied to the point at any step of this technique, so the present embodiment may work with other SPR functions in conjunction with noting areas of moving text or HSF areas.

[0020] If the point is detected as an edge of text or HSF areas, then a current edge counter is incremented to count the number of edge points in a frame (thus, it may be desirable to reset the counter to zero at the beginning of each frame) at step 412. At step 414, the location of every current n-th edge point is storedpossibly in storage 306where n is selected to give a good performance statistically to note areas of moving text or HSF areas. The number n takes on all possible ranges between 1 and the total number of addressable points on the screen. However, n1 (i.e. save up to every possible addressable point on the screen) may be a useful metric if the system designer would want near perfect information as to where all edges of HSF text and images are locatedbut a lesser number of points would suffice to give a good indication that there are HSF areas in motion on the screen. With nthe total number of addressable points on screen (i.e. save one point of information every screen), this may not be useful as a metric as there may not be enough good data to indicate where there are significant amount of moving HSF text and images to warrant taking an action. Thus, the number n is optimally in between these two extreme values.

[0021] It will be appreciated that other embodiments could have other criteria for selecting and storing locations of points, including random selection. It is not necessary that a data is stored in modulo arithmetic fashion. It suffices that there are a sufficient number of points to note moving text and HSF areas. At step 416, the image data is incremented to the next location and begins processing at step 402 until there is an end of frame condition detected.

[0022] FIG. 5 continues with the processing at step 406/502 with a comparison of current edge points against the same points in the previous frame's stored values. If the comparison indicates that the current edge points are the same as the previous frame's edge points, then the test indicates that there is little or no motion detected and to go process the next frame of data and to turn on SPR as necessary (for example, if SPR had been turned off via this processing previously) at step 504. If, on the other hand, the comparison indicates that the current edge points are different enough from the previous frame's edge points, then motion is detectedenough so that the system can take corrective action at step 506such as turning off SPR for the next frame (and other successive frames as necessary until the motion has stopped).

[0023] It should be appreciated that same and different encompass many possible metrics. Same could mean that not one edge point has changed (or has been added or deleted) from one frame to the next. Alternatively, same could mean that the system tolerates up to a certain number or a certain percentage of edge changes without need for taking corrective action. Also, the system might even take into consideration a percentage change in a certain subset area of the screen as either same or different. One embodiment might consider that a certain percentage change in a smaller subset area of the screen means that there is a high possibility that there is a window opened (e.g. word processor) that does not take up the full screen and that HSF information is moving. In such a case, the system might turn off SPR for that portion of the screen and leave the remaining screen as previously treated. Of course, the level of same and different could be preset into the system according to either a heuristic analysis or an empirical data analysis. These levels may be considered threshold levels or values and may be dynamically controlled by an artificial intelligent mechanism or alternatively, be set by the users themselves.

[0024] It should also be appreciated that the current frame and the previous frame may not necessarily be successive frames. It may suffice that the current and previous frame have a relevant relationship (i.e. every other frame or the like, or two related frames in an MPEG format) that might serve as a basis of noting that motion is being detected. Additionally, instead of comparing individual points frame by frame, if there is a MPEG encoding (or some other suitable encoding), it may be possible to detect changes in motion vectors.

[0025] At step 508, the current frame's edge data is transferred to the previous frame's data and the system is effectively reset at step 510 (e.g., the edge counter and current memory location for storing edge data can be reset) and ready to process another frame's worth of image

[0026] It will be appreciated that there are many possible embodiments and variations on the above embodiments and notions. It would suffice for the purposes of the present invention that the system be able to detect that HSF image data is in motion and that, if there is a level of such motion detected that wouldin the estimation of the systemdetract from the user's perspective (i.e. too much color error introduced), then the system can take corrective actions such as turn off subpixel rendering for all or a portion of the screen or effectively alter the SPR in some way to correct the viewer's experience of the image. Instead of turning off SPR, the SPR can also be changed to another filter that is less sensitive to motion artifacts of the LCD. In fact, some of the alternative corrective actions are described in the two related patent applications noted above and incorporated herein and in the other patent applications also incorporated herein.

[0027] Another alterative way of describing this technique is as follows:

[0028] At beginning of frame, set counter to zero.

[0029] Increment counter every time an edge is detected.

[0030] Store the location of every nth edge detection in a temporary memory.

[0031] Compare number of edges detected to a preset number.

[0032] If number of edges exceeds preset number, then set a flag indicating text is present.

[0033] Next frame, repeat process AND check to see if edges are stationary i.e. locations are the same of the statistical sample of edges.

[0034] If flag is set AND locations are not the same, then motion is detected. In this case, turn off all SPR on edges (when adaptive filter returns true value). Color error would be observed, but since the text is moving, is can not be easily seen. The rest of the image will be unchanged; for example if text is moving only in a small region of the screen.

[0035] If flag is not set AND locations are not the same, then SPR is on normally.

[0036] If flag is set AND locations are the same, then SPR is on normally.

[0037] If flag is not set AND locations are the same, then SPR is on normally.

[0038] Repeat.

[0039] A simplification (as shown in FIGS. 6 and 7) to the above process is to just compare the number of edges from frame to frame in memories Mem1 and Mem2 that can store count values; if the number is different by a preset amount, the text is moving (as long as the number of edges exceed the minimum). Even in the case of the text being turned off, there will be little disturbance in image quality since the action will be detected the next frame (₁/60^th second later) and the flag will be set to off. The size of the temporary memory (the value of n) will depend on the accuracy required for this process. This can be determined empirically.

[0040] Referring to FIG. 6, a counter for counting edge data points is reset at step 602. A data point is inputted at step 604. The data point is tested at step 606 to determine if the data point is at the end of a frame. If the data point is at the end of the frame, the process continues to step 608 of FIG. 7. If the data is at the end of the frame, the data point is tested at step 610 to determine if it is an edge data point. If not, the process continues from step 610 back to step 604. If yes, the counter is incremented at step 612.

[0041] At step 614 of FIG. 7 that continues from step 608 of FIG. 6, the total number edges (i.e., the count value from the counter) is stored in a memory (Mem2). This count value can provide the total number of edges for a current frame of input data. Another memory (Mem1) can store the total number edges for a previous frame of input data. At step 616, a test is made to determine if the total number of edges in Mem2 is different from the total number of edges in Mem1. At step 620, if the total number is not different (i.e., if Mem2Mem10), then no motion is detected and the number of edges in Mem1. At step 618, if the total number is different, motion is detected and different SPR can be applied used on the edge data points and the new value can be stored in Mem1.

[0042] As a refinement to all embodiments, the SPR could be altered on only the text or edges of HSF that areas moving and not to edges that are not moving. One embodiment for accomplishing this task is when moving edges are detected, the graphical subsystem can send a query back to the operating system to enquire as to what active windows might be open that would have moving HSF edges (e.g. word processors, image writers, etc). If there is an application having such an open window, the graphical subsystem could either ask the operating system and/or application to suspend any subpixel rendering mode for its image data inside the window or ask the operating system and/or application to give the dimensions of such window and the graphical subsystem would then alter or shut off SPR for those dimensions on screen.

[0043] An alternative embodiment that would not need to talk to the operating system might be for the graphical subsystem to turn off (or otherwise alter) SPR for all edges within a certain neighborhood of edges that are detected as moving. In this manner, most moving edges would have their SPR altered to substantially correct the color error introduced by movement. In such a case, it would be desirable to have a sufficiently large number of edges stored for comparison so that desirable subsets of the screen (i.e. scrolling windows) would be shut off or suitably altered.

[0044] It has now been disclosed several embodiments of the techniques, systems and methods of the present invention. It should be appreciated the many variations on these embodiments are possible and that the scope of the present invention is not limited to the embodiments disclosed herein; but encompasses the many possible variations to the same.