US 7,970,208 B2Grant

Apparatus to detect homogeneous region of image using adaptive threshold

Issue Date:2011-06-28

•9 Claims

•4 Drawing Sheets

Abstract

A homogeneous region detector of an image using an adaptive threshold, and a method of the same. The homogeneous region detector includes a global region standard deviation calculation part to calculate a global region standard deviation of a whole region of an input image, a local region standard deviation calculation part to divide the input image into a predetermined number of local regions and to calculate a local region standard deviation of the each local region, and a homogeneous region determination part to separate the homogeneous region from a feature region in the input image using an adaptive threshold calculated based on entropy of the input image. Accordingly, the homogeneous region can be precisely detected by effectively separating the homogeneous region from the feature region of the input image by using the threshold adaptively calculated based on the entropy of the input image, so that the homogeneous region can be applied to various fields of image processing.

Metadata

Assignee

Samsung Electronics Co., Ltd.

Inventor

Seung-hoon Han

Application Information

Application Number:US 11/237,942

Filing Date:2005-09-29

Priority Date:2004-12-28

Art Unit:2624

Classifications

IPC:

G06K9/34G06K9/38

Patent Drawings (4 sheets)

Description

Cross-Reference to Related Applications

[0001] This application claims benefit under 35 U.S.C. §119(a) of Korean Patent Application No. 2004-113795, filed Dec. 28, 2004, the disclosure of which is incorporated herein in its entirety by reference.

Background of the Invention

[0002] 1. Field of the Invention

[0003] The present general inventive concept relates to an apparatus to detect a homogeneous region of an image, and a method thereof. More particularly, the present general inventive concept relates to a homogeneous region detector capable of correctly detecting a homogeneous region of an image using an adaptive threshold appropriate to properties of the image, and a method thereof.

[0004] 2. Description of the Related Art

[0005] Generally, a homogeneous region in an image is used for a variety of image processing fields. For example, the homogeneous region may be used for estimation of image noise since a signal-to-noise (S/N) ratio of the homogeneous region is low. Also, due to spatial redundancy of the homogeneous region, efficiency of image compression can be improved using the homogeneous region. Furthermore, when detecting scene transition of video, the homogeneous region enhances motion tolerance of frame differences. For image interpolation, an interpolation value can be obtained thorough a simple calculation in the homogeneous region, thereby saving costs. As explained above, the homogeneous region is applied to diverse fields of image processing. Therefore, correct detection of the homogeneous region is an important matter.

[0006] FIG. 1 is a view illustrating a conventional method for detecting a homogeneous region of an image. Referring to FIG. 1, a conventional homogeneous region detector 100 comprises a local region standard deviation calculation part 10 and a comparison part 20. The local region standard deviation calculation part 10 divides an input image into M×M regions and calculates the standard deviation of each region. The comparison part 20 compares the standard deviation calculated with respect to the each region to a preset fixed threshold T. A region having a smaller standard deviation than the preset fixed threshold T is determined to be a homogeneous region.

[0007] However, with the conventional detecting method using the fixed threshold, it is hard to precisely detect the homogeneous region according to an image or image noise. For example, a texture region could be detected as a homogeneous region, or a real homogeneous region could fail to be detected as the homogeneous region according to the fixed threshold. When detection of the homogeneous region is not correct, noise estimation in the detected region or other post-processing of the image cannot be effectively performed.

Summary of the Invention

[0008] The present general inventive concept provides a homogeneous region detector capable of correctly detecting a homogeneous region of an image, using an adaptive threshold calculated based on entropy of the image, and a method thereof.

[0009] Additional aspects of the present general inventive concept will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the general inventive concept.

[0010] The foregoing and/or other aspects of the present general inventive concept may be achieved by providing a homogeneous region detector using an adaptive threshold, the homogeneous region detector including a global region standard deviation calculation part to calculate a global region standard deviation of a whole region of an input image, a local region standard deviation calculation part to divide the input image into a certain number of local regions and to calculate a local region standard deviation of each local region, and a homogeneous region determination part to separate a homogeneous region from a feature region in the input image using the adaptive threshold calculated based on entropy of the input image.

[0011] The homogeneous region determination part may include an adaptive threshold calculation part to calculate the adaptive threshold based on the global region standard deviation and an entropy constant K of the input image, and a comparison part to compare the adaptive threshold with the local region standard deviation of each local region and to determine one of the local regions to be the homogeneous region when the adaptive threshold is greater than the respective local region standard deviation.

[0012] The adaptive threshold may be calculated by the following expression:

[0013]

T 1 = K \frac{σ_{g}}{\log 2 σ_{g}}

[0014] wherein, T1 denotes the adaptive threshold, K denotes the entropy constant, and σ_gdenotes the global region standard deviation of the input image.

[0015] The entropy constant may refer to bit/symbol.

[0016] The foregoing and/or other aspects of the present general inventive concept may also be achieved by providing a homogeneous region detector to detect a homogeneous region of an input image, the homogenous region detector including a calculation unit to calculate a global standard deviation of the input image, to calculate a plurality of local standard deviations corresponding to a plurality of local regions of the input image, and to calculate an adaptive threshold based on the calculated global standard deviation and an entropy constant of the input image, and a comparison unit to compare the plurality of local standard deviations to the adaptive threshold calculated by the calculation unit to determine whether the corresponding plurality of local regions of the input image are homogeneous.

[0017] The foregoing and/or other aspects of the present general inventive concept may also be achieved by providing a homogenous region detector to detect a homogenous region of an input image, the homogenous region detector including an adaptive threshold calculation unit to calculate an adaptive threshold according to a predetermined property of the input image, a standard deviation calculation unit to divide the input image into a plurality of regions and to calculate a local standard deviation of each region, and a comparison unit to compare the local standard deviation of each region to the adaptive threshold calculated by the adaptive threshold calculation part to determine a homogenous region of the input image.

[0018] The foregoing and/or other aspects of the present general inventive concept may be achieved by providing a method of detecting a homogeneous region of an image using an adaptive threshold, the method including calculating a global region standard deviation of a whole region of an input image, dividing the input image into a certain number of local regions and calculating a local region standard deviation of each local region, and separating a homogeneous region from a feature region in the input image using the adaptive threshold calculated based on entropy of the input image.

[0019] The separating of the homogeneous region from the feature region may include calculating the adaptive threshold based on the global region standard deviation and an entropy constant of the input image, and comparing the adaptive threshold with the local region standard deviation of each local region and determining one of the local regions to be the homogeneous region when the adaptive threshold is greater than the respective local region standard deviation.

[0020] The adaptive threshold may be calculated by the following expression:

[0021]

T 1 = K \frac{σ_{g}}{\log 2 σ_{g}}

[0022] wherein, T1 denotes the adaptive threshold, K denotes the entropy constant, and σ_gdenotes the global region standard deviation of the input image.

[0023] The entropy constant may refer to bit/symbol.

[0024] The foregoing and/or other aspects of the present general inventive concept may also be achieved by providing a method of detecting a homogenous region of an input image, the method including calculating an adaptive threshold according to a predetermined property of the input image, calculating a plurality of local standard deviations corresponding to a plurality of regions of the input image, and comparing the plurality of local standard deviations to the calculated adaptive threshold to determine a homogenous region of the input image.

Brief Description of the Drawings

[0025] These and/or other aspects of the present general inventive concept will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

[0026] FIG. 1 is a view illustrating a conventional method for detecting a homogeneous region of an image;

[0027] FIG. 2 is a block diagram illustrating a homogeneous region detector using an adaptive threshold, according to an embodiment of the present general inventive concept; and

[0028] FIG. 3 is a flowchart illustrating a method of detecting a homogeneous region of an image using an adaptive threshold, according to an embodiment of the present general inventive concept.

Detailed Description of the Preferred Embodiments

[0029] Reference will now be made in detail to the embodiments of the present general inventive concept, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout. The embodiments are described below in order to explain the present general inventive concept while referring to the figures.

[0030] FIG. 2 illustrates a homogenous region detector 200 usable with an image processing apparatus according to an embodiment of the present general inventive concept. Referring to FIG. 2, the homogeneous region detector 200 includes a global region standard deviation calculation part 210, a local region standard deviation calculation part 220 and a homogeneous region determination part 230.

[0031] The global region standard deviation calculation part 210 calculates a global region standard deviation σ_gof an input image. The global region standard deviation σ_gobtained by the global region standard deviation calculation part 210 is input to the homogeneous region determination part 230 to be used to detect a homogeneous region.

[0032] The local region standard deviation calculation part 220 divides the input image into M×M regions and calculates a local region standard deviation σ_mof each region. The local region standard deviation σ_mobtained by the local region standard deviation calculation part 220 corresponding to each region of the input image is input to the homogeneous region determination part 230 to be used to detect the homogeneous region.

[0033] The homogeneous region determination part 230 comprises an adaptive threshold calculation part 240 and a comparison part 250. The adaptive threshold calculation part 240 calculates an adaptive threshold T1 based on entropy of the input image. That is, the adaptive threshold calculation part 240 calculates the adaptive threshold T1 on the basis of the global region standard deviation σ_gand an entropy constant K of the image. Here, the entropy constant K refers to bit/symbol. That is, the entropy constant K can be a measure of information per symbol in the input image. The comparison part 250 compares the adaptive threshold T1 calculated by the adaptive threshold calculation part 240 with the local region standard deviation σ_mof each region of the input image to determine the homogeneous region of the input image. When the local region standard deviation σ_mof one of the regions of the input image is less than the adaptive threshold T1, the comparison part 240 determines that the region is a homogeneous region.

[0034] FIG. 3 is a flowchart illustrating a method of detecting a homogeneous region of an image using an adaptive threshold according to an embodiment of the present general inventive concept. Referring to FIGS. 2 and 3, the global region standard deviation σ_gof the input image and the local region standard deviation σ_mof each region of the input image are respectively calculated by the global region standard deviation calculation part 210 and the local region standard deviation calculation part 220 (S310). The global region standard deviation σ_gof the input image and the local region standard deviation σ_mof each region of the input image are then supplied to the homogeneous region determination part 230

[0035] The adaptive threshold calculation part 240 in the homogeneous region determination part 230 calculates the adaptive threshold T1 based on the global region standard deviation σ_gand the entropy constant K of the input image (S320). The adaptive threshold T1 is calculated by the adaptive threshold calculation part 240 as follows.

[0036] A probability distribution P_gof a global region of the input image with respect to brightness and a probability distribution P_mof the homogeneous region with respect to brightness can be calculated from the global region standard deviation σ_g. Assuming that the global region standard deviation has a wide range, the probability distribution P_gof the global region of the input image with respect to the brightness can be approximated to a uniform density function as illustrated in [Expression 1] below.

[0037]

\begin{matrix} P_{g} (x) ≅ \frac{1}{2 σ_{g}}, - σ_{g < x <} σ_{g} & [Expression 1] \end{matrix}

[0038] wherein, P_gdenotes the probability distribution of the global region of the input image, σ_gdenotes the global region standard deviation of the input image, and x denotes a pixel value of each pixel of the input image.

[0039] Assuming that a standard deviation of the homogeneous region has a very narrow range, the probability distribution P_mof the homogeneous region with respect to the brightness can be approximated to the following expression.

[0040]

\begin{matrix} P_{m} (x) ≅ \frac{1}{2 σ_{g}}, - σ_{m < x <} σ_{m} & [Expression 2] \end{matrix}

[0041] The entropy of each local region is calculated based on the probability distributions P_g(x) and P_m(x) of the global and homogeneous regions obtained through [Expression 1] and [Expression 2], respectively, as illustrated in [Expression 3] below.

[0042]

\begin{matrix} H (m) = - \sum_{x = σ_{g}}^{σ_{g}} P_{m} \log P_{m} ≅ \frac{σ_{m}}{σ_{g}} \log 2 σ_{g} & [Expression 3] \end{matrix}

[0043] In [Expression 3], H(m) denotes the entropy of each local region as divided into M×M regions. When the local region entropy H(m) is an entropy constant K, it can be expressed as the following.

[0044]

\begin{matrix} H (m) ≅ \frac{σ_{m}}{σ_{g}} \log 2 σ_{g} \approx K & [Expression 4] \end{matrix}

[0045] Thus, assuming that the local region entropy H(m) is substantially the same as the entropy constant K, when the global region standard deviation σ_gand log 2σ_gare transposed to a term including the entropy constant K, only the local region standard deviation σ_mremains on the left term of the expression. The right term can now be expressed as the following.

[0046]

\begin{matrix} K \frac{σ_{g}}{\log 2 σ_{g}} & [Expression 5] \end{matrix}

[0047] Referring to [Expression 5], assuming that the local region entropy H(m) is substantially the same as the entropy constant K, the adaptive threshold T1 can be finally obtained by transposing the global region standard deviation σ_gand log 2σ_gto the term including the entropy constant K. That is the adaptive threshold T1 can be set to be equal to the maximum local region standard deviation σ_mof the homogeneous region, which can be obtained by solving for the local region standard deviation σ_min [Expression 4]. Accordingly, the adaptive threshold T1 is

[0048]

K \frac{σ_{g}}{\log 2 σ_{g}}

as shown in [Expression 5].

[0049] When the adaptive threshold T1 is calculated at operation S320, the comparison part 250 compares the local region standard deviation am of each region of the input image with the adaptive threshold T1 (S330). When the local region standard deviation σ_mcorresponding to one of the regions of the input image is less than the adaptive threshold T1 as a result of the comparison (S340), the corresponding local region is determined to be the homogeneous region (S350). When the local region standard deviation is not less than the adaptive threshold T1 as a result of the comparison (S340), the corresponding local region is determined to be not the homogeneous region (S360).

[0050] The present general inventive concept may be embodied as executable code in computer readable media including storage media such as magnetic storage media (ROMs, RAMs, floppy disks, magnetic tapes, etc.), and optically readable media (CD-ROMs, DVDs, etc.), and carrier waves (transmission over the Internet).

[0051] As described above, according to an embodiment of the present general inventive concept, the homogeneous region of an image can be correctly detected by using a threshold adaptively calculated according to entropy of the image.

[0052] Furthermore, a homogeneous region and a feature region of an image, which is not homogeneous, can be effectively separated by using a threshold adaptively calculated according to entropy of the image, thereby precisely detecting the homogeneous region of the image. Accordingly, a homogeneous region can be applied to diverse fields of image processing.

[0053] According to an embodiment of the present general inventive concept, since a homogeneous region is determined using only a global region standard deviation and a local region standard deviation, calculation processes can be simplified.

[0054] Although a few embodiments of the present general inventive concept have been shown and described, it will be appreciated by those skilled in the art that changes may be made in these embodiments without departing from the principles and spirit of the general inventive concept, the scope of which is defined in the appended claims and their equivalents.

Claims

What is claimed is:

1. An apparatus to detect homogeneous region of an input image, the apparatus comprising:

a global region standard deviation calculation part to calculate a global region standard deviation of a whole region of the input image;

a local region standard deviation calculation part to divide the input image into a predetermined number of local regions and to calculate a local region standard deviation of each local region; and

a homogeneous region determination part to separate a homogeneous region from a non-homogeneous region in the input image using an adaptive threshold calculated based on an entropy constant of the input image and the global region standard deviation,

wherein the homogeneous region determination part comprises:

an adaptive threshold calculation part to calculate the adaptive threshold based on the global region standard deviation and the entropy constant of the input image; and

a comparison part to compare the calculated adaptive threshold with the local region standard deviation of each local region and to determine one of the local regions to be the homogeneous region when the adaptive threshold is greater than the respective local region standard deviation,

wherein the adaptive threshold calculation part calculates the adaptive threshold assuming that a local region entropy calculated based on the adaptive threshold, the global region standard deviation and the local region standard deviation is the same as the entropy constant.

2. An apparatus to detect homogeneous region of an input image, the apparatus comprising:

a global region standard deviation calculation part to calculate a global region standard deviation of a whole region of the input image;

a local region standard deviation calculation part to divide the input image into a predetermined number of local regions and to calculate a local region standard deviation of each local region; and

wherein the adaptive threshold is calculated by the following expression:

T 1 = K \frac{σ_{g}}{\log 2 σ_{g}}

wherein T1 denotes the adaptive threshold, K denotes an entropy constant of the input image, and σ_gdenotes the global region standard deviation of the input image.

3. The apparatus of claim 2, wherein the entropy constant K comprises a bit/symbol measurement of the input image.

4. A method to detect a homogeneous region of an image using an adaptive threshold in a homogenous region detector, the method comprising:

calculating a global region standard deviation of a whole region of an input image, dividing the input image into a certain number of local regions, and calculating a local region standard deviation of each local region; and

separating a homogeneous region from a non-homogeneous region in the input image using an adaptive threshold calculated based on an entropy constant of the input image and the global region standard deviation,

wherein the above steps are computed by a digital computer,

wherein the separating of the homogenous region from the non-homogenous region in the image comprises:

calculating the adaptive threshold based on the global region standard deviation and the entropy constant of the input image; and

comparing the calculated adaptive threshold with the local region standard deviation σ_mof each local region and determining one of the local regions to be the homogeneous region when the adaptive threshold is greater than the respective local region standard deviation, and

wherein an adaptive threshold calculation part calculates the adaptive threshold assuming that a local region entropy calculated based on the adaptive threshold, the global region standard deviation and the local region standard deviation is the same as the entropy constant.

5. A method to detect a homogeneous region of an image using an adaptive threshold in a homogenous region detector, the method comprising:

wherein the above steps are computed by a digital computer, and

wherein the adaptive threshold is calculated by the following expression:

T 1 = K \frac{σ_{g}}{\log 2 σ_{g}}

wherein, T1 denotes the adaptive threshold, K denotes an entropy constant of the input image, and σ_gdenotes the global region standard deviation of the input image.

6. The method of claim 5, wherein the entropy constant comprises a bit/symbol measurement of the input image.

7. A computer readable recording medium comprising executable codes to perform a method of detecting a homogenous region of an image, the method comprising:

wherein the separating of the homogenous region from the non-homogeneous region in the image comprises:

calculating the adaptive threshold based on the global region standard deviation and the entropy constant of the image; and

8. The apparatus of claim 2, wherein the homogeneous region determination part comprises:

an adaptive threshold calculation part to calculate the adaptive threshold based on the global region standard deviation and the entropy constant of the input image; and

9. The method of claim 5, wherein the separating of the homogenous region from the non-homogeneous region in the image comprises:

calculating the adaptive threshold based on the global region standard deviation and the entropy constant of the input image; and

Patent Citations (28)

Patent	Date	Inventor	Cited By
US4567610(A)	1986-01-01	McConnell	Examiner
US4741046(A)	1988-04-01	Matsunawa et al.	Examiner
US5027413(A)	1991-06-01	Barnard	Examiner
US5204738(A)	1993-04-01	Aragaki	Examiner
US5915039(A)	1999-06-01	Lorie et al.	Examiner
US5949905(A)	1999-09-01	Nichani et al.	Examiner
US6021221(A)	2000-02-01	Takaha	Examiner
US6347156(B1)	2002-02-01	Kamada et al.	Examiner
US6870956(B2)	2005-03-01	Qi et al.	Examiner
US6970578(B1)	2005-11-01	Strand	Examiner
US6993169(B2)	2006-01-01	Wetzel et al.	Examiner
US7330600(B2)	2008-02-01	Nishida	Examiner
US7394938(B2)	2008-07-01	Erol et al.	Examiner
US7474790(B2)	2009-01-01	Dalal et al.	Examiner
US7653242(B2)	2010-01-01	Haque	Examiner
US2002/0028008(A1)	2002-03-01	Fan et al.	Examiner
US2002/0172412(A1)	2002-11-01	Jun et al.	Examiner
US2003/0174887(A1)	2003-09-01	Oh et al.	Examiner
US2004/0008892(A1)	2004-01-01	Hill et al.	Examiner
US2004/0076337(A1)	2004-04-01	Nishida	Examiner
US2005/0163374(A1)	2005-07-01	Ferman et al.	Examiner
US2007/0160286(A1)	2007-07-01	Haque	Examiner
JP2002-92609	1999-09-01		Applicant
JP2004-72446	2004-01-01		Applicant
KR1996-27399	1996-07-01		Applicant
KR98-72846	1998-11-01		Applicant
KR1999-69865	1999-09-01		Applicant
KR2004-46271	2004-06-01		Applicant

Non-Patent Literature (5)

Trier and Jain, “Goal-Directed Evaluation of Binarization Methods”, IEEE 1995, p. 47-58.Examiner
Sahoo, P et al. “Threshold selection using a minimal histogram entropy difference”, Optical Engineering 36, 1997 p. 1976-1981.Examiner
Navon et al. “Color image segmentation based on adaptivve local thresholds”, Image and Vision Computing 23, 2004, p. 69-85.Examiner
Gonzalez, Rafael and Woods, Richard, “Digital Image Processing”, 2002, Prentice Hall, Second Edition, pp. 437-440, 600-606.Examiner
Sezgin, Mehmet and Sankur, Bulent, “Survey over image thresholding techniques and quantitative performance evaluation”, Jan. 2004, Journal of Electronic Imaging 13(1), pp. 146-168.Examiner

Abstract

Description

Cross-Reference to Related Applications

Background of the Invention

[0002] 1. Field of the Invention

[0004] 2. Description of the Related Art

Summary of the Invention

[0012] The adaptive threshold may be calculated by the following expression:

[0013]

T 1 = K \frac{σ_{g}}{\log 2 σ_{g}}

[0014] wherein, T1 denotes the adaptive threshold, K denotes the entropy constant, and σ_gdenotes the global region standard deviation of the input image.

[0015] The entropy constant may refer to bit/symbol.

[0020] The adaptive threshold may be calculated by the following expression:

[0021]

T 1 = K \frac{σ_{g}}{\log 2 σ_{g}}

[0022] wherein, T1 denotes the adaptive threshold, K denotes the entropy constant, and σ_gdenotes the global region standard deviation of the input image.

[0023] The entropy constant may refer to bit/symbol.

Brief Description of the Drawings

[0026] FIG. 1 is a view illustrating a conventional method for detecting a homogeneous region of an image;

[0027] FIG. 2 is a block diagram illustrating a homogeneous region detector using an adaptive threshold, according to an embodiment of the present general inventive concept; and

[0028] FIG. 3 is a flowchart illustrating a method of detecting a homogeneous region of an image using an adaptive threshold, according to an embodiment of the present general inventive concept.

Detailed Description of the Preferred Embodiments

[0037]

\begin{matrix} P_{g} (x) ≅ \frac{1}{2 σ_{g}}, - σ_{g < x <} σ_{g} & [Expression 1] \end{matrix}

[0040]

\begin{matrix} P_{m} (x) ≅ \frac{1}{2 σ_{g}}, - σ_{m < x <} σ_{m} & [Expression 2] \end{matrix}

[0042]

\begin{matrix} H (m) = - \sum_{x = σ_{g}}^{σ_{g}} P_{m} \log P_{m} ≅ \frac{σ_{m}}{σ_{g}} \log 2 σ_{g} & [Expression 3] \end{matrix}

[0044]

\begin{matrix} H (m) ≅ \frac{σ_{m}}{σ_{g}} \log 2 σ_{g} \approx K & [Expression 4] \end{matrix}

[0046]

\begin{matrix} K \frac{σ_{g}}{\log 2 σ_{g}} & [Expression 5] \end{matrix}

[0048]

K \frac{σ_{g}}{\log 2 σ_{g}}