CN1797471A - Method for detecting area of skin color of human body in image compression domain - Google Patents

Abstract

The invention discloses a method for detecting human body skin color area in JPEG image compression domain. The method includes the following steps: initializing system parameters; performing Huffman decoding and inverse quantization on the JPEG image to be detected to obtain Y, Cb, Cr color components DCT coefficients corresponding to each image block; calculate the color feature and texture feature of each image block; calculate the skin color probability of each image block; judge whether each image block is a human skin color area according to the skin color probability and texture feature. The invention has the advantages that: the method of the invention does not need to completely decompress the image, and directly operates on the compressed code stream, thereby improving the detection speed of the human body skin color area and improving the detection accuracy of the human body skin color area.

Description

Human skin color region detection method in image compression domain

technical field

The invention relates to a method for detecting a human skin color area in an image, in particular to a detection method for a human body skin color area in a JPEG image compression domain.

Background technique

An important type of information in images is human information, and human skin is an important physiological feature of human beings. Fast and accurate detection of human skin color regions in images has important theoretical and practical value in applications such as face detection and sensitive image filtering.

JPEG (Joint Photographic Experts Group) is an international still image compression standard formulated by the still image compression standard formulation committee under the International Organization for Standardization (ISO). Due to the excellent quality of JPEG for image static compression, it has achieved great success. Eighty percent of the images on the website currently use the JPEG compression standard.

Due to the wide applicability of JPEG, when detecting human skin color, it is unavoidable to require skin color area detection on images compressed by JPEG standard. In the existing skin color area detection method, for images compressed by the JPEG standard, the main steps of skin color detection are:

1) Huffman decoding is performed on the JPEG image to be detected to obtain the DCT coefficient of each image block;

2) Perform inverse DCT decoding on each image block, decompress the image into the pixel domain, and realize the complete decompression of the image;

3) Perform skin color detection and discrimination pixel by pixel on the fully decompressed image.

The above is just a brief description of the existing skin color area detection method. For more information about the skin color area detection method, please refer to Reference 1: In 1999 at the Computer Vision and Pattern Recognition conference, the paper "Statistical Color Models with application to skin detection".

Existing detection methods need to fully decompress the image into the pixel domain before performing skin color detection. In the process of complete decompression, the computational complexity of inverse DCT decoding for each image block is high, and the detection process is pixel by pixel. However, only the color information is considered, and the texture information is not considered, which leads to a large amount of calculation required for skin color detection and low detection accuracy.

If the operation steps of the existing human skin color detection method are simplified and texture information is fully considered, the calculation amount of skin color detection can be reduced and the detection accuracy can be improved.

Contents of the invention

The purpose of the present invention is to provide a human body skin color area detection method on the JPEG image compression domain, which overcomes the defects of complex operation steps and large calculation amount in the existing detection method, and realizes rapid detection of human body skin color area.

Yet another object of the present invention is to provide a method for detecting human skin color areas in the JPEG image compression domain, which overcomes the defect of low detection accuracy of skin color areas in existing detection methods, and realizes accurate detection of human body skin color areas.

In order to achieve the above object, the invention provides a method for detecting human body skin color regions on the JPEG image compression domain, the method may further comprise the steps:

a) Initialize system parameters, system parameters include: texture feature threshold, skin color probability threshold, image block size, human skin color distribution model;

b) Huffman decoding and inverse quantization are carried out on the JPEG image to be detected, and the JPEG image to be detected is decomposed into image blocks to obtain Y, Cb, DCT coefficients corresponding to each image block on the Cr color component;

c) using the DCT coefficients obtained in step b) to calculate the color features and texture features of each image block;

d) using the color feature obtained in step c) and the human skin color distribution model set in step a), to calculate the skin color probability of each image block;

e) Judging whether each image block is a human skin color area according to the skin color probability and texture features.

In the above technical solution, in step a), the size of the image block is one of the following three: 8×8, 4×4, 2×2.

In the above technical solution, in step c), the values of the color features of the image block on the Y, Cb, and Cr three color components are respectively the DC coefficients of the DCT coefficients of the image block on the corresponding color components Divide by the square root of the number of pixels in the pixel block of the image block.

In the above technical solution, in step c), the value of the texture feature of the image block is the sum of the squares of the non-zero AC coefficients in the DCT coefficients on the Y color component of the pixel block divided by the number of pixels in the pixel block.

In the above technical solution, in step e), the criterion for judging that the image block is a human skin color area is: the skin color probability of the image block is within the range of the set skin color probability threshold and the texture feature is within the set texture feature threshold. within range.

The advantage of the inventive method is:

1. The method of the present invention does not need to fully decompress the image, and directly operates on the compressed code stream, which simplifies the operation steps and improves the detection speed of the human body skin color area.

2. In the detection process, the texture information of the detection image is fully considered, which improves the detection accuracy of the human skin color area.

Description of drawings

Fig. 1 is the flowchart of the inventive method

Figure 2 is a schematic diagram of the DCT coefficients of a 4×4 image block

Detailed ways

The method of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

As shown in Figure 1, the human skin color area detection method on the JPEG image compression domain of the present invention mainly comprises the following steps:

In step 100, system parameters are initialized. These parameters include: texture feature threshold, skin color probability threshold, image block size, human skin color distribution model. The texture feature threshold and color feature threshold are fixed values obtained by experiments. The image block size refers to the size of the basic image block when detecting the human skin color area, the image block size should be an integral power of 2, and must be one of the following three: 8×8, 4×4, 2×2; Human skin color distribution model depicts the possibility of various colors belonging to human skin color. Various existing or future human skin color distribution models are applicable to the method of the present invention, such as statistical skin color model and single Gaussian model.

In step 200, the JPEG image to be detected is subjected to Huffman decoding and dequantization, and the JPEG image to be detected is decomposed into image blocks to obtain DCT coefficients corresponding to each image block on Y, Cb, and Cr color components.

In a color JPEG image, each color is generally represented by three color components, namely: Y, Cb, Cr. Therefore, each N×N image block corresponds to an image block on the three color components. The value of each color component of each N×N size image block undergoes a discrete cosine transform (DCT) to obtain a corresponding discrete cosine transform (DCT) coefficient, and the coefficient is a matrix of N×N size. The first coefficient of this matrix is the direct current (DC) coefficient, and the other coefficients are called alternating current (AC) coefficients. As shown in Figure 2, the figure is a 4×4 DCT coefficient, where the first coefficient (0, 0) in the matrix is a direct current (DC) coefficient, and the other coefficients in the matrix are alternating current (AC) coefficients .

With reference to the JPEG image compression standard, the DCT coefficients corresponding to each 8×8 image block can be obtained by Huffman decoding; in step 100, there are three possibilities for the size of the image block, if the image block size set in step 100 is 4×4, each 8×8 image block needs to be decomposed into four 4×4; if the image block size set in step 100 is 2×2, each 8×8 image block needs to be decomposed into 16 2 ×2 image blocks. The process of decomposing the 8×8 image block is realized by operating on DCT, that is, according to the DCT coefficient of the original 8×8 image block, the DCT coefficient of each 4×4 or 2×2 image sub-block is calculated.

Taking an 8×8 image block as an example, the decomposition of the image block will be described.

Note that the DCT coefficients corresponding to 8×8 blocks are D88; the corresponding DCT coefficients of four 4×4 blocks are D44 ₁ , D44 ₂ , D44 ₃ , D44 ₄ ; the corresponding DCT coefficients of 16 2×2 blocks are D22 ₁ , D22 ₂ , D22 ₃ , D22 ₄ , D22 5, D22 ₆ , D22 ₇ , D22 ₈ , _{D22 9} , D22 ₁₀ , D22 ₁₁ , D22 ₁₂ , D22 ₁₃ , D22 ₁₄ , D22 ₁₅ , D22 ₁₆ . D88 is an 8×8 matrix, D44 ₁ to D44 ₄ are all 4×4 matrices, and D22 ₁ to D22 ₁₆ are all 2×2 matrices.

The method of decomposing an 8×8 block into four 4×4 blocks is as follows:

$\left[\begin{array}{cc}\mathrm{<span\; class="notranslate">\; D.</span>}{\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; D.</span>}{\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 84</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 88</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 84</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

where D84 is the transformation matrix and D84 ^T is its transpose matrix. D84 is a fixed value, its value is:

$\left[\begin{array}{cccccccc}\mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.4531</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.1591</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1063</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.0901</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.2079</span>}& \mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.3955</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.1762</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1389</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.0373</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.2566</span>}& \mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.3841</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.1877</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.011</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.0488</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.2452</span>}& \mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.4329</span>}\\ \mathrm{<span\; class="notranslate">\; 0.5</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.4531</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1591</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.1063</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.0901</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.2079</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.3955</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.1762</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1389</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.0373</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.2566</span>}& \mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.3841</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1877</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.0114</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.0488</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.2452</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.4329</span>}\end{array}\right]$

The method of decomposing 4×4 blocks into 2×2 blocks is as follows:

$\left[\begin{array}{cc}\mathrm{<span\; class="notranslate">\; D.</span>}{\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}{\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

$\left[\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

$\left[\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 9</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 10</span>}}\\ {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 11</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 12</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}{\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

$\left[\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 13</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 14</span>}}\\ {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 15</span>}}& {\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}_{\mathrm{<span\; class="notranslate">\; 16</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; D.</span>}{\mathrm{<span\; class="notranslate">\; 44</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 42</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}$

where D42 is the transformation matrix and D42 ^T is its transpose matrix.

D42 is a fixed value, its value is:

$\left[\begin{array}{cccc}\mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.4619</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.1913</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1913</span>}& \mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.4619</span>}\\ \mathrm{<span\; class="notranslate">\; 0.5</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.4619</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1913</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0.1913</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}& \mathrm{<span\; class="notranslate">\; 0.4619</span>}\end{array}\right]<span\; class="notranslate">\; .</span>$

From 8×8 blocks to 16 2×2 blocks, you can first decompose 8×8 blocks into 4 4×4 blocks, and then decompose each 4×4 block into 4 2×2 blocks.

In step 300, the color features and texture features of each image block are calculated.

The color feature of the image block is represented by the YCbCr color space, and its values on the three color components of Y, Cb, and Cr are respectively the DC coefficient of the DCT coefficient of the image block on the corresponding color component divided by the pixel block of the image block The square root of the number of pixels of is expressed as: $\mathrm{the\; y}=\frac{1}{N}{C}_{\mathrm{the\; y}}\left(\mathrm{0,0}\right),$ $\mathrm{cb}=\frac{1}{N}{C}_{\mathrm{cb}}\left(\mathrm{0,0}\right),$ $\mathrm{cr}=\frac{1}{N}{C}_{\mathrm{cr}}\left(\mathrm{0,0}\right).$

The texture feature of the image block is the sum of the squares of the non-zero AC coefficients in the DCT coefficients on the Y color component of the pixel block divided by the number of pixels in the pixel block, that is $\mathrm{texture}=\frac{1}{N^2}\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{\left(C_{\mathrm{the\; y}}(i,j)\right)}^2,(i,j)\&NotEqual;\left(\mathrm{0,0}\right),$ In the above formula, y, cb, and cr are the values of the color features of the image block on the three color components of Y, Cb, and Cr, respectively, and texture is the texture feature of the image block. C _y , C _cb , and C _cr are the DCT coefficients of the image block on the Y, Cb, and Cr components respectively, and the size of each image block is N×N, that is, there are N ² pixels.

In step 400, the likelihood of skin color is calculated for each image patch. The value of the color feature of the image block is obtained in step 300, the skin color probability of each color value can be known from the skin color distribution model in step 100, the value of the color feature of the image block is compared with each color value in the skin color distribution model, and the image is obtained The skin color probability of the block.

In step 500, each image block is judged according to the skin color probability and texture features, and whether it is a human skin color area is judged. The criterion for judging that the image block is a human skin color image block is that the skin color probability of the image block is within the range of the set skin color probability threshold and the texture feature is within the set texture feature threshold range.

Claims (5)

1, a human body skin color region detection method on the JPEG image compression domain, the method may further comprise the steps: a) Initialize system parameters, system parameters include: texture feature threshold, skin color probability threshold, image block size, human skin color distribution model; b) Huffman decoding and inverse quantization are carried out on the JPEG image to be detected, and the JPEG image to be detected is decomposed into image blocks to obtain Y, Cb, DCT coefficients corresponding to each image block on the Cr color component; c) using the DCT coefficients obtained in step b) to calculate the color features and texture features of each image block; d) using the color feature obtained in step c) and the human skin color distribution model set in step a), to calculate the skin color probability of each image block; e) Judging whether each image block is a human skin color area according to the skin color probability and texture features. 2, the human skin color area detection method on the JPEG image compression domain according to claim 1, is characterized in that, in described step a), the value of described image block size is one of following three: 8×8, 4×4, 2×2. 3. The human body skin color area detection method on the JPEG image compression domain according to claim 1, characterized in that, in step c), the color features of the image block are on Y, Cb, and Cr three color components The values of are respectively the square root of the DC coefficient in the DCT coefficient of the image block on the corresponding color component divided by the number of pixels in the pixel block of the image block. 4. The human body skin color region detection method on the JPEG image compression domain according to claim 1, characterized in that, in step c), the value of the texture feature of the image block is the DCT on the Y color component of the pixel block The sum of the squares of the non-zero AC coefficients in the coefficients is divided by the number of pixels in the pixel block. 5. The human body skin color area detection method on the JPEG image compression domain according to claim 1, characterized in that, in the step e), the criterion for judging that an image block is a human body skin color area is: the skin color of the image block The probability is within the range of the set skin color probability threshold and the texture feature is within the set texture feature threshold.

Images