CN1761205A - System for detecting eroticism and unhealthy images on network based on content - Google Patents

Abstract

The system includes following sub systems: icon detection subsystem filters out images on network in small size and strip shape from advertisement of network sites; text detection subsystem determines text image and suspicious image; color detection sub system analyzes color compositions of image, builds model of skin color and degree of exposure in order to separate normal image from suspicious image; gesture detection sub system builds feature base of erotic standard image to determine comparability matched to erotic images. The invention takes the lead in applying search technique of image recognition based on content to filtering erotic images on Internet. The invention raises filtering effect, providing identifying and filtering success ratio more than 99% and misjudging ratio less than 5%.

Description

Content-based network pornography image and bad image detecting system

One, technical field: the present invention relates to the Internet flame filtration system, particularly relate to a kind of content-based network pornography image and bad image detecting system.

Two, background technology: the Internet is worldwide popularized rapidly as a kind of modernized communication technology, and the Internet communication approach spreads all over each corner in the world.Because network world is a Virtual Space, all data in real life, sound, information such as image can change computer bit kenel into, shuttle back and forth in the whole world with computerized information stream, at present increasing people is engaged in amusement on network, research and commercial activity, thereby form a virtual society on the network, network user need not disclose the true identity of oneself and can enjoy a trip to therebetween, the interpersonal also rare morals of daily society, the ethics constraint, therefore network world is more complicated more than real society, fearful, panoramic personage is mingled with therebetween, nourish different purposes separately, justice, evil difficulty is distinguished.Because ordering about of violence, porn site and pornographic webpage are mad in recent years increases, particularly the strong harmful informations such as pornographic image of stimulus to the sense organ are overflowed, bring out juvenile deliquency, have a strong impact on pupillary growing up healthy and sound, cause the head of a family's very big indignation and worry, also cause the concern of society and government, even send the cry of " helping child ", causing spends huge sums set up in, primary school's campus gateway has closed the passage of leading to network education, also disabled for the household PC that child buys, wasted investment greatly, have to give up online superior educational resource.

For online harmful informations such as filtering eroticisms, a large amount of filter softwares and system have also appearred on the market in recent years, can be referred to as " blacklist software ", its technological means is with artificial means known pornographic network address or domain name to be included in " blacklist " address base, by address comparison and keyword comparison, listed network address and relevant information in " blacklist " that the blockade viewer lands.The shortcoming of this method is: for a large amount of undiscovered powerless with pornographic network address that increase newly and the conversion looks, the discovery of intelligence that can not be real-time also is included into blacklist, and the literal comparison time also is subjected to the restriction of country variant literal, is in passive filtration state all the time.A kind of Sexy file judging system and the method for Chinese invention patent ZL0112132.7, elder generation's input marking file in system, word segment in the isolated tested webpage and picture part are sent to literal relatively engine and porny identification engine respectively, also compare by the pornographic index that calculates literal and picture, judge Sexy file with pornographic index of discrimination; A kind of Sexy picture checking system of ZL0112127.0 filters the examine picture by dual engine, has introduced relatively engine of porny database and database, has improved the accuracy of porny identification; A kind of Web content filtration system of patent application 200410053683.3, by information filtering agency, querying server and content analysis and management server, the information filtering proxies store has blacklist and white list, querying server has one and has the URL storehouse of rating information suddenly of classifying, content analysis and management server are classified and classified estimation to the resource among the internet, system has self-learning capability, the genealogical classification precision can be improved, all kinds of media datas that exist in the Internet can be initiatively filtered; Its main filtration approach remains based on automatic renewal and interception to URL, lack profound, completely, based on the filtration of flesh and blood, a large amount of strong pornographic image of stimulus to the sense organ can not directly be tackled still arranged.

Three, summary of the invention:

Technical problem to be solved by this invention: at the defective that the present the Internet of background technology pornographic image detects, filtration system exists, propose a kind of content-based, multi-level the Internet pornographic image and bad image detecting system, set up characteristic model and 100,000 the pornographic image standard feature storehouses of oneself.

The technical solution adopted in the present invention:

A kind of content-based network pornography image and bad image detecting system, mainly contain skin color detection subsystem and attitude detection subsystem, system sets up the Mathematical Modeling of skin color detection and attitude detection fast algorithm, the color detection subsystem is formed by the skin color of phase-split network image and the experiment in color of image space is compared, adopt the hsv color space to set up complexion model, the skin color of determining the people is in selected hsv color spatial distributions situation, and then computed image colour of skin degree of exposure, determine a threshold values of differentiating image colour of skin degree of exposure, distinguish normal picture and suspect image in view of the above; Described attitude detection subsystem, at first pick out the representative standard pornographic image of some, after carrying out signature analysis, extract its feature and set up the posture feature storehouse by training, it is pornographic standard picture feature database, whether as judgement is the foundation of the coupling similitude of pornographic image, by the suspect image on the network is carried out Wavelet Edge Detection, obtain an edge image, by the Wavelet Edge image is analyzed, extract marginal point, determine the boundary rectangle of object, pixel in the rectangle is cut apart according to complexion model, the skin area image of tentatively being cut apart is converted to gray level image then through the morphologic filtering corrosion treatment, by shape description and the posture analysis to the skin area image, image in the pornographic characteristics of image of present image and the standard storehouse is mated similar judgment processing, definition similarity d _i, set thresholding T_shape, obtain N characteristic similarity d _iAfter, if characteristic similarity drops on interval [T_shape, 1], then think feature similarity in present image feature and the feature database, and the number Num of statistics similar features, if Num satisfies condition: Num＞T_num, wherein T_num is the threshold value of N feature similarity number in present image feature and the feature database, think that so this image is a pornographic image, otherwise adjudicating this image is normal picture.

By described skin color detection subsystem, at first the pixel transitions with network image is the hsv color space and quantizes, be divided into L color sub-spaces, determine the total shin_count and the frequency sub_count_i of sample skin pixels in this L sub spaces of sample skin pixels then by statistical analysis, wherein satisfy i=1, Λ, L $\underset{i=1}{\overset{L}{\mathrm{\Σ}}}\mathrm{sub}\_\mathrm{count}\_i=\mathrm{shin}\_\mathrm{count}$

Be distributed in the possibility of this subspace with the normalized frequency as skin pixels,

v _i＝sub_count_i/skin_count

Set the possibility threshold value T_vi of a colour of skin distribution probability, if satisfy v _i〉=T_vi, then w _i=v _iOtherwise, w _i=0; Final like this obtaining: A={A ₁, A ₂, Λ, A _L}

W＝{w ₁，w ₂，Λ，w _L}

Wherein, w _iRepresent corresponding subspace A _iDegree of membership, i.e. A _iIn color be the possibility of skin color, i=1,2, Λ L, parameter L gets 72, cluster obtains the degree of membership set W of the distribution subspace set A of skin color and A; Computed image colour of skin degree of exposure:

To arbitrary image F (x, y), x=1, Λ, M, y=1, Λ, N, (x y) is transformed into hsv color space and quantification, obtains this color of pixel subspace label, makes entire image F (x with each pixel, y) just changed into a M * N label dot matrix G (m, n), statistics G (m, n) normalization histogram Hue[k], k=1, Λ, L is by the colour of skin degree of exposure in the following formula computed image $\mathrm{Ratio}=\underset{k=1}{\overset{L}{\mathrm{\Σ}}}\mathrm{Hue}\left[k\right]{\×w}_k$

Utilize image colour of skin degree of exposure Ratio to distinguish normal picture and pornographic image then, take two kinds of judgement modes: (1) hard decision: determine a threshold value T_Valve, relatively Ratio and T_Valve adjudicate: if piece image satisfies Ratio 〉=T_Valve, then adjudicating this image is pornographic image; Otherwise be normal picture, the value of T_Value is taken between [0.10,0.15]; (2) soft-decision: determine a low threshold value T_Low, a high threshold T_High, relatively Ratio and these two threshold values are adjudicated: if piece image satisfies Ratio 〉=T_High, then adjudicating this image is pornographic image; If satisfy Ratio≤T_Low, then adjudicating this image is normal picture; Think under other situations that this image is a suspect image, this detector is not done judgement, and the attitude detection subsystem that passes on detects;

Described attitude detection subsystem, the attitude detection core algorithm mainly contains Wavelet Edge Detection, image segmentation, morphologic filtering, shape description and similarity and mates several parts:

Wavelet Edge Detection adopts the Daubechies-4 wavelet basis that the suspicious original image on the network is carried out tower wavelet decomposition, obtains LL low frequency sub-band and LH, HL, and three high-frequency sub-band of HH are utilized following formula

$E[i,j]={(E_1{[i,j]}^2+E_2{[i,j]}^2+E_3{[i,j]}^2)}^{\frac{1}{2}}$

To three types of edges synthesize an edge graph E (i, j);

Image segmentation, at first the Wavelet Edge image is analyzed, extracted four marginal points in upper and lower, left and right, and determine the boundary rectangle of object according to this, wipe then and be positioned at the outer pixel of boundary rectangle in the original color image, pixel in the rectangle is cut apart according to complexion model, to any pixel p (x, y), it is transformed into the HSV space and quantizes to obtain quantizing label k ∈ [1, Λ, L], if w _k≠ 0, then keep this pixel, otherwise, wipe this pixel, the skin area image of tentatively being cut apart;

Morphologic filtering adopts mathematical morphology that the image of tentatively cutting apart is handled, and filters out the noise pixel that does not belong to object area;

Shape description, after obtaining the area image of object, utilize the second order of image and 7 constant Hu squares that third moment can draw image:

φ＝η ₂₀+η ₀₂

${\mathrm{\φ}}_2={({\mathrm{\η}}_{20}-{\mathrm{\η}}_{02})}^2+4{\mathrm{\η}}_{11}^2$

φ ₃＝(η ₃₀-3η ₁₂) ²+(3η ₂₁-η ₀₃) ²

φ ₄＝(η ₃₀+η ₁₂) ²+(η ₂₁+η ₀₃) ²

φ ₅＝(η ₃₀-3η ₁₂)(η ₃₀+η ₁₂)[(η ₃₀+η ₁₂) ²-3(η ₀₃+η ₂₁) ²]

+(3η ₂₁-η ₀₃)(η ₂₁+η ₀₃)[3(η ₃₀+η ₁₂) ²-(η ₀₃+η ₂₁) ²]

φ ₆＝(η ₂₀-η ₀₂)[(η ₃₀+η ₁₂) ²-(η ₂₁+η ₀₃) ²]+4η ₁₁(η ₃₀+η ₁₂)(η ₂₁+η ₀₃)

φ ₇＝(3η ₂₁-η ₀₃)(η ₃₀+η ₁₂)[(η ₃₀+η ₁₂) ²-3(η ₀₃+η ₂₁) ²]

+(3η ₁₂-η ₃₀)(η ₂₁+η ₀₃)[3(η ₃₀+η ₁₂) ²-(η ₀₃+η ₂₁) ²]

Adopt 7 characteristic values of 18 characteristic values of second order to five rank normalization central moment of image and Hu square to describe a width of cloth and cut apart later skin area feature of image shape;

The similarity coupling adopts weighting Euclidean distance to carry out measuring similarity, and establishing weight vector is W _j, present image is characterized as φ _j, j=1 wherein, 2, K, 25; Feature database is characterized as φ _Ij', i=1,2, K, N, j=1,2, K, 25, wherein N representation feature Al Kut is levied number, definition similarity d _iFor

$d_i=1-{\left(\underset{j=1}{\overset{25}{\mathrm{\Σ}}}{W}_j{({\mathrm{\φ}}_j-{\mathrm{\φ}}_{\mathrm{ij}}^{\′})}^2\right)}^{\frac{1}{2}}$

Obtain N characteristic similarity d _iAfter, set thresholding T_shape, if characteristic similarity drops on interval [T_shape, 1], then think feature similarity in present image feature and the feature database, and the number Num of statistics similar features, if Num satisfies condition: Num＞T_num, wherein T_num is the threshold value of N feature similarity number in present image feature and the feature database, thinks that so this image is a pornographic image, otherwise adjudicating this image is normal picture.

Described network pornography image and bad image detecting system, also contain the icon detection subsystem, according to the size of images ratio network image is differentiated, at first to the width and the height setting threshold values T-size of image, judge according to the size of network image then, filtering out less than this setting threshold values, be the too little bad network image that is generally icon one class of size, is normal picture greater than the then judgement of this setting threshold values; Secondly, judge, set the ratio threshold values T-logo of picture altitude and width according to the ratio of the height and the width of image, filter out laterally or most longitudinally be the network image of the fillet shape of advertiser web site and so on, the T-size value selects 32, the T-logo value selects 10.

Described network pornography image and bad image detecting system, also contain the text detection subsystem, according to text image and the general difference of continuous-tone image on color is formed, by to the histogrammic analysis of color of image, choose suitable gray value as dividing histogrammic threshold values, H[i], i ∈ [0,255], get θ _Eg〉=200 are divided into low gray value and two zones of high gray value as thresholding with grey level histogram, utilize following formula to calculate the energy proportion in high gray value zone: $p_{\mathrm{eg}}=\underset{{i=\mathrm{\θ}}_{\mathrm{gg}}}{\overset{255}{\mathrm{\Σ}}}H\left[i\right]/\underset{i=0}{\overset{255}{\mathrm{\Σ}}}H\left[i\right],$ To satisfy P _Eg〉=P _EGImage be judged as text image, according to identification requirement P _EGCan choose different values, generally choose P _EG〉=0.7; Perhaps different with the comentropy that general continuous-tone image is shown according to text image, choose certain gray value range Theta _Ep1≤ i≤θ _Ep2, calculate its histogram information entropy, select θ _Ep1=127, θ _Ep2=255, histogram is done normalized: $P\left[i\right]=H\left[i\right]/\underset{i=0}{\overset{255}{\mathrm{\Σ}}}H\left[i\right],$ Compute histograms local message entropy: ${\mathrm{ep}}_l=-\underset{i={\mathrm{\θ}}_{\mathrm{ep}1}}{\overset{{\mathrm{\θ}}_{\mathrm{ep}2}}{\mathrm{\Σ}}}P\left[i\right]\log P\left[i\right],$ To satisfy ep _l〉=EP _LImage be judged as text image, require EP according to identification _LDesirable different value is generally got EP for text image _L≤ 2; Perhaps differentiate the result of text image, above-mentioned two kinds of methods are carried out fusion treatment: P according to colouring information _EgSelected threshold P _EG1And P _EG2And to ep _lSelected threshold EP _L1And EP _L2, then definition:

$\mathrm{EG}=\left\{\begin{array}{cc}0,& p_{\mathrm{eg}}<P_{\mathrm{EG}1};\\ \frac{p_{\mathrm{eg}}-P_{\mathrm{EG}1}}{P_{\mathrm{EG}2}-P_{\mathrm{EG}1}}& P_{\mathrm{EG}1}\&le;p_{\mathrm{eg}}<P_{\mathrm{EG}2};\\ 1& p_{\mathrm{eg}}\&GreaterEqual;P_{\mathrm{EG}2};\end{array}\right.$

$\mathrm{EP}=\left\{\begin{array}{cc}0,& {\mathrm{ep}}_l>{\mathrm{EP}}_{L2};\\ 1-\frac{{\mathrm{ep}}_l-{\mathrm{EP}}_{L1}}{{\mathrm{EP}}_{L2}-{\mathrm{EP}}_{L1}}& {\mathrm{EP}}_{L1}<{\mathrm{ep}}_l\&le;{\mathrm{EP}}_{L2};\\ 1& {\mathrm{ep}}_l\&le;{\mathrm{EP}}_{L1};\end{array}\right.$

Definition is based on the text image identification parameter of color:

$C_H=\frac{\mathrm{EG}+\mathrm{EP}}{2}$

C _H∈ [0,1]; Then work as C _HTo look like be text image to decision diagram during greater than a certain threshold parameter T_ch.

Described network pornography image and bad image detecting system, contain other bad image detection subsystem, the feature samples of other specific bad image is carried out the PCA conversion in rgb color space, set up the PCA color space, in conjunction with neural net to the colour of skin sample training in the PCA color space, obtain a stable characteristics detector, the suspect image that obtains through icon detector and text detector by and the comparison of this property detector, detect bad network image and be input to the color detection subsystem and be determined further processing.

Described network pornography image and bad image detecting system by differentiating the similar ratio of characteristic image coupling in webpage pornographic image and the pornographic standard picture feature database, are provided with the pornographic image rank.

Described network pornography image and bad image detecting system, systems soft ware is embedded on the high-speed dsp image detection card hard card of parallel processing, described hard card contains digital signal processing circuit and pci bus interface circuit, digital signal processor adopts TMS320C6711 on the hard card, the synclk circuit is connected with the corresponding port of electrification reset Dongle circuit and central processing unit, the SDRAM external memory storage is connected by the I/O port of bus interface and central processing unit with flash memory FLASH, the host interface of central processing unit connects CPLD and programmable logic device, programmable logic device is selected PLX9054 for use, perhaps select PLX9052 for use, perhaps adopt the S5920 of AMCC, or S5933.

Positive beneficial effect of the present invention:

1, the present invention takes the lead in being applied to the detection filtration aspect of the Internet pornographic image with technology with " content-based image recognition retrieval " is theoretical at home, created content-based bad image detection model, in conjunction with cluster and neural net method, merged the icon detection, multi-level Intelligent Measurement technology such as text detection and pornographic image, network address by past passive is filtered the information filtering that jumps to active, improved filter effect significantly, can filter JPAG, GIF, BMP, the various picture formats of TIF, the integral body of the Internet pornographic image is discerned filtering success rate greater than 99%, False Rate is lower than 5%, to other flame filter effect greater than 80%, the average recognition time of pornographic image less than 0.5 second, is not influenced networking speed.

2, pornographic image detection model of the present invention, through contrast screening repeatedly, set up 100,000 in the pornographic characteristics of image of standard storehouse, as judging that whether network image is the foundation of the similitude judgement of pornographic image, has realized that content-based flame filters detection, can directly tackle pornographic image information, and in real time pornographic network address being added blacklist automatically, the real-time update url database is in it all the time and dynamically updates, have intelligent, interception efficient height.

Four, description of drawings:

Fig. 1: content-based network pornography image and bad image detecting system are formed block diagram

Fig. 2: based on the bad image training and the testing flow chart of neural net and PCA conversion

Fig. 3: network pornography image and bad image detection model overall structure pattern and application flow

Five, embodiment:

Embodiment one: referring to Fig. 1, Fig. 3, network pornography image and bad image detecting system contain the icon detection subsystem, by the size of images ratio network image are differentiated, purpose is to detect the image of those similar advertiser web sites, filters out too little image simultaneously.Since these images be rendered as mostly one very narrow rectangular, perhaps the size of integral image is smaller, does not generally constitute harm from content.

(1) differentiates according to the size of image:, think to belong to icon one class less than the image of this threshold value to the width and the height setting threshold value of image.

Min (image_width, image_height)＜T_size, then judgement is normal picture.

(2) differentiate according to the ratio of the height and the width of image: the proportion threshold value of setting height and width, can screen fillet image laterally or longitudinally like this, they generally mostly are advertiser web site and so on.

if(image_width＞image_height)R _s＝image_width/image_height；

elseR _s＝image_height/image_width。

If (R _s＞T_logo), then judgement is normal picture.

In practice, rule of thumb, our selected threshold T_size=32, T_logo=10.

Described network pornography image and bad image detecting system, contain the text detection subsystem, network image is carried out text/image to be differentiated, text detector is carried out text/image to network image and is differentiated, the image that detection is made up of large amount of text information, for example fax through internet that exists with image format, network character advertisement etc.

(1) histogram is divided

By to the histogrammic analysis of color of image, find that character image and continuous-tone image have a great difference, have such characteristics: in the higher regional centralized of gray value most energy, and on remaining gray scale, be similar to even distribution.According to these characteristics, choose suitable gray value as dividing histogrammic threshold value, identify character image according to the contrast of the energy of tonal range before and after it.

To input picture, be converted into gray level image earlier, a kind of simple way is promptly got the brightness value of each picture element.Statistics obtains the histogram of this gray level image, H[i], i ∈ [0,255].According to a large amount of experiments, get θ _Ep〉=200 are divided into low gray value and two zones of high gray value as thresholding with grey level histogram.Utilize following formula to calculate the energy proportion in high gray value zone:

$p_{\mathrm{eg}}=\underset{{i=\mathrm{\&theta;}}_{\mathrm{gg}}}{\overset{255}{\mathrm{\&Sigma;}}}H\left[i\right]/\underset{i=0}{\overset{255}{\mathrm{\&Sigma;}}}H\left[i\right]$

To satisfy P _Eg〉=P _EGImage be judged as text image, P here _EGThe desirable different value that requires according to identification.For text image, test shows, gets P _EGThe 〉=0.7th, suitable.

(2) local message entropy: color character image wants dull a lot of relatively because the color of continuous-tone image is abundant, thereby the comentropy difference that both showed, and histogrammic local message entropy is with difference performance more obvious of the two.Choose the gray value range Theta _Ep1≤ i≤θ _Ep2, calculate its histogram information entropy, select θ here _Ep1=127, θ _Ep2=255, histogram is done normalization:

$P\left[i\right]=H\left[i\right]/\underset{i=0}{\overset{255}{\mathrm{\&Sigma;}}}H\left[i\right]$

Compute histograms local message entropy: ${\mathrm{ep}}_l=-\underset{i={\mathrm{\&theta;}}_{\mathrm{ep}1}}{\overset{{\mathrm{\&theta;}}_{\mathrm{ep}2}}{\mathrm{\&Sigma;}}}P\left[i\right]\log P\left[i\right]$

To satisfy ep _l〉=EP _LImage be judged as text image, EP here _LEqually according to the desirable different value of discerning that requires.For text image, get EP _LThe≤2nd, suitable.

(3) fusion treatment

Result as foundation colouring information differentiation text image can merge the result that above-mentioned two kinds of methods obtain.Method is as follows: to p _EgSelected threshold P _EG1And P _EG2And to ep _lSelected threshold EP _L1And EP _L2, then definition:

$\mathrm{EG}=\left\{\begin{array}{cc}0,& p_{\mathrm{eg}}<P_{\mathrm{EG}1};\\ \frac{p_{\mathrm{eg}}-P_{\mathrm{EG}1}}{P_{\mathrm{EG}2}-P_{\mathrm{EG}1}}& P_{\mathrm{EG}1}\&le;p_{\mathrm{eg}}<P_{\mathrm{EG}2};\\ 1& p_{\mathrm{eg}}\&GreaterEqual;P_{\mathrm{EG}2};\end{array}\right.$

$\mathrm{EP}=\left\{\begin{array}{cc}0,& {\mathrm{ep}}_l>{\mathrm{EP}}_{L2};\\ 1-\frac{{\mathrm{ep}}_l-{\mathrm{EP}}_{L1}}{{\mathrm{EP}}_{L2}-{\mathrm{EP}}_{L1}}& {\mathrm{EP}}_{L1}<{\mathrm{ep}}_l\&le;{\mathrm{EP}}_{L2};\\ 1& {\mathrm{ep}}_l\&le;{\mathrm{EP}}_{L1};\end{array}\right.$

Definition is based on the text image identification parameter of color:

$C_H=\frac{\mathrm{EG}+\mathrm{EP}}{2}$

C _H∈[0，1]。Then work as C _HTo look like be text image to decision diagram during greater than a certain threshold parameter T_ch.

The skin color detection subsystem; Color by the phase-split network image forms and the experiment in color of image space is compared; Adopt the hsv color space to set up complexion model; Determine that the Person's skin color is in the distribution situation in selected hsv color space; At first the pixel transitions with network image is the hsv color space and quantizes; Be divided into L color sub-spaces; Then determine total shin_count and the frequency sub_count_i of sample skin pixels in this L sub spaces of sample skin pixels by statistical analysis; Wherein satisfy i=1; Λ; L

$\underset{i=1}{\overset{L}{\mathrm{\&Sigma;}}}\mathrm{sub}\_\mathrm{count}\_i=\mathrm{shin}\_\mathrm{count}$

Be distributed in the possibility of this subspace with the normalized frequency as skin pixels,

v _i＝sub_count_i/skin_count

Set the possibility threshold value T_vi of a colour of skin distribution probability, if satisfy v _i〉=T_vi, then w _i=v _iOtherwise, w _i=0; Final like this obtaining: A={A ₁, A ₂, Λ, A _LW={w ₁, w ₂, Λ, w _L}

Wherein, w _iRepresent corresponding subspace A _iDegree of membership, i.e. A _iIn color be the possibility of skin color, i=1,2, Λ L, parameter L=72, cluster obtains the degree of membership set W of the distribution subspace set A of skin color and A;

Computed image colour of skin degree of exposure: to arbitrary image F (x, y), x=1, Λ, M, y=1, Λ, N, (x y) is transformed into hsv color space and quantification, obtains this color of pixel subspace label, makes entire image F (x with each pixel, y) just changed into a M * N label dot matrix G (m, n), statistics G (m, n) normalization histogram Hue[k], k=1, Λ, L is by the colour of skin degree of exposure in the following formula computed image

$\mathrm{Ratio}=\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}\mathrm{Hue}\left[k\right]\&times;w_k$

Utilize image colour of skin degree of exposure Ratio to distinguish normal picture and pornographic image then, take two kinds of judgement modes: (1) hard decision: determine a threshold value T_Valve, relatively Ratio and T_Valve adjudicate: if piece image satisfies Ratio 〉=T_Valve, then adjudicating this image is pornographic image; Otherwise be normal picture, the value of T_Value is taken between [0.10,0.15]; (2) soft-decision: determine a low threshold value T_Low, one high

Threshold value T_High, relatively Ratio and these two threshold values are adjudicated: if piece image satisfies Ratio 〉=T_High, then adjudicating this image is pornographic image; If satisfy Ratio≤T_Low, then adjudicating this image is normal picture; Think under other situations that this image is a suspect image, this detector is not done judgement, and the attitude detection subsystem that passes on detects;

The attitude detection subsystem is at first set up the posture feature storehouse by training, to carrying out posture analysis and similar coupling by the suspect image of color detector, distinguishes normal picture and pornographic image.The gesture detector algorithm mainly is made up of several parts such as Wavelet Edge Detection, image segmentation, morphologic filtering, shape description and similarity couplings, and each several part specifically describes as follows:

(1) Wavelet Edge Detection

Traditional Wavelet Edge Detection principle is: establish C _J+1Represent original image, C _j, D _j ¹, D _j ², D _j ³Be raw video through four width of cloth subimages that wavelet transformation obtains, establish ({ h _k} _{K ∈ Z}, { g _k} _{K ∈ Z)}With $({\left\{{\tilde{h}}_k\right\}}_{k\&Element;Z},{\left\{{\tilde{g}}_k\right\}}_{k\&Element;Z})$ Be one group of dual filter that biorthogonal wavelet is derived, then decomposition of the biorthogonal wavelet of image and reconstruction formula are as follows:

$\left\{\begin{array}{c}{C}_{j,m,n}=\underset{k,j\&Element;Z}{\mathrm{\&Sigma;}}{C}_{j+1,k,l}{h}_{k-2m}{h}_{l-2n}\\ D_{j,m,n}^1=\underset{k,j\&Element;Z}{\mathrm{\&Sigma;}}{C}_{j+1,k,l}{h}_{k-2m}{h}_{l-2n}\\ D_{j,m,n}^2=\underset{k,j\&Element;Z}{\mathrm{\&Sigma;}}{C}_{j+1,k,l}{h}_{k-2m}{h}_{l-2n}\\ D_{j,m,n}^3=\underset{k,j\&Element;Z}{\mathrm{\&Sigma;}}{C}_{j+1,k,l}{h}_{k-2m}{h}_{l-2n}\end{array}\right.$

$C_{j+1,m,n}=(\underset{k,l\&Element;Z}{\mathrm{\&Sigma;}}{C}_{j,k,l}{\tilde{h}}_{m-2k}{\tilde{h}}_{n-2l}+\underset{k,l\&Element;Z}{\mathrm{\&Sigma;}}{D}_{j,k,l}^1{\tilde{h}}_{m-2k}{\tilde{g}}_{n-2l}$

$+\underset{k,l\&Element;Z}{\mathrm{\&Sigma;}}{D}_{j,k,l}^2{\tilde{g}}_{m-2k}{\tilde{h}}_{n-2l}+\underset{k,l\&Element;Z}{\mathrm{\&Sigma;}}{D}_{j,k,l}^3{\tilde{g}}_{m-2k}{\tilde{g}}_{n-2l})$

The detected image marginal point is promptly sought in certain neighborhood along the gradient vector direction and is made that the gradient vector amplitude is the point of maximum so, and the gradient vector amplitude is proportional to:

$D_j=\sqrt{{\left|D_j^1\right|}^2+{\left|D_j^2\right|}^2}$

And the direction vector of this gradient is: Arg (D _j ¹+ iD _j ²).

In application, as fruit dot (x, gradient vector amplitude D y) _jBe the local maximum point in the neighborhood on the direction vector of this gradient, satisfy simultaneously: D _j＞T, T are thresholding, and then this point is considered to marginal point.

We adopt the Daubechies-4 wavelet basis that original image is carried out tower wavelet decomposition, obtain LL low frequency sub-band and LH, HL, three high-frequency sub-band of HH.Wherein, the LH subband comprises the edge on the original image horizontal direction; The HL subband comprises the edge on the original image vertical direction; The HH subband comprises the edge on the original image diagonal.We detect as above three types edge respectively, and three types of edges that obtain are synthesized an edge graph.The LH subband is sought gradient vector amplitude maximum point in certain neighborhood in the horizontal direction, and the wavelet coefficient that only keeps the LH subband carries out inverse wavelet transform, obtains edge subgraph E ₁(i, j).Similar HL subband and HH subband are handled, obtained E respectively ₂(i, j) and E ₃(i, j) edge subgraph.Utilize following formula to three types of edges synthesize an edge graph E (i, j).

$E[i,j]={(E_1{[i,j]}^2+E_2{[i,j]}^2+E_3{[i,j]}^2)}^{\frac{1}{2}}$

Image by the skin color detector is a coloured image, and we handle gray level image when carrying out Wavelet Edge Detection often, therefore coloured image can be converted to gray level image earlier or directly utilize the red channel of coloured image to handle.

(2) image segmentation for the shape to object in the image is described, is cut apart image in conjunction with Wavelet Edge image and complexion model, mainly therefrom is partitioned into the human body complexion area exposed.

At first, the Wavelet Edge image is analyzed, extract the most left, the rightest, go up most, the most following four marginal points, and determine the boundary rectangle of object with this; Then, wipe the pixel that is positioned in the original color image outside the object boundary rectangle.Pixel in the rectangle is cut apart according to complexion model.(x y), is transformed into it HSV space and quantizes to obtain quantizing label k ∈ [1, Λ, L] to any pixel p.If w _k≠ 0, then keep this pixel; Otherwise, wipe this pixel.The skin area image of tentatively being cut apart.

(3) morphologic filtering

The skin area image of tentatively cutting apart that produces above often exists very little graininess of a lot of areas and spot shape noise, need carry out Filtering Processing to them, filter out the noise pixel that those do not belong to object area, effectively keep those pixels that belong to object area simultaneously.Filtering method commonly used as low pass, high pass, level and smooth etc., at this, adopts mathematical morphology to come the image of tentatively cutting apart is handled.

Morphology has defined four kinds of basic operations such as expansion, burn into unlatching, closure, and wherein unlatching and closure operation are the compound operations of expansion and erosion operation.For input picture f, the setting structure element is b, and f and b are image in essence, and then b is defined as the expansion of f

$(f\&CirclePlus;b)\left(s\right)=\max \{f(s-x)+b\left(x\right)|x\&Element;D_b,\&Exists;(s-x)\&Element;D_f\}$

B is defined as the corrosion of f

(fΘb)(s)＝min{f(s+x)-b(x)|x∈D _b，(s+x)∈D _f}

B is defined as the unlatching of f

fob＝(fΘb)b

B is defined as the closure of f

f·b＝(fb)Θb

Wherein, D _fAnd D _bBe respectively the domain of definition of f and b, s and x are integer Z ²Vector in the space.For dilation operation,, can expand as long as structural element b and input picture f have a pixel to intersect.On the contrary,, have only when structural element b all is positioned at f, just can corrode for erosion operation.Go up expansion energy expansion image aspects, and corrosion energy downscaled images form from how much.Open computing and can remove the convex domain that does not match with structural element on the image, keep the convex domain that those match simultaneously.Closure operation is then filled the recessed zone that does not match with structural element on those images, keeps the recessed zone that those match simultaneously.To the skin area image of tentatively cutting apart, adopt the morphological erosion operator to handle, to the image behind the erosion operation, be converted into gray level image earlier, carry out region description then.

(4) shape description: after obtaining the area image of object, shape how to describe this width of cloth image has various ways, describes as digital metric, Fourier description, square description and the topology of region shape.Haveing nothing to do because the translation of Hu square and image, rotation and engineer's scale change, is very useful to the shape description of image.Totally 25 characteristic values of 7 characteristic values that we adopt 18 characteristic values of second order to five rank normalization central moment of image and Hu square are described a width of cloth and are cut apart later skin area feature of image shape.

(5) similarity coupling: adopt weighting Euclidean distance to carry out measuring similarity.If weight vector is W _j, current image feature is φ _j, j=1 wherein, 2, K, 25; Feature database is characterized as φ _Ij', i=1,2, K, N, j=1,2, K, 25, wherein N representation feature Al Kut is levied number.Definition similarity d _iFor

$d_i=1-{\left(\underset{j=1}{\overset{25}{\mathrm{\&Sigma;}}}{W}_j{({\mathrm{\&phi;}}_j-{\mathrm{\&phi;}}_{\mathrm{ij}}^{\&prime;})}^2\right)}^{\frac{1}{2}}$

Obtain N characteristic similarity d _iAfter, set thresholding T_shape, if characteristic similarity drops on interval [T_shape, 1], then think feature similarity in present image feature and the feature database, and add up the number Num of similar features.If Num satisfies condition: Num＞T_num, wherein T_num is the threshold value of N feature similarity number in current image feature and the feature database, thinks that so this image is a pornographic image.Otherwise adjudicating this image is normal picture.

Network pornography image of the present invention and bad image detecting system, contain other bad image detection subsystem, referring to Fig. 2, other bad image detection subsystem, the feature samples of other specific bad image is carried out the PCA conversion in rgb color space, set up the PCA color space, in conjunction with neural net to the colour of skin sample training in the PCA color space, obtain a stable characteristics detector, the suspect image that obtains through icon detector and text detector by and the comparison of this property detector, detect bad network image and be input to the color detection subsystem and be determined further processing.Other bad visual detector and pornographic image detector concept are similar, but the image recognition of counterpart's body characteristics, and bad image lacks the feature of general character, therefore can only adopt the pattern of training, comparison to adjudicate.Under many circumstances, people are transformed into HSI space or YCbCr space with rgb color space, and monochrome information is separated with chrominance information, utilize the HS two-dimensional sub-spaces in the HSI space or the CbCr two-dimensional sub-spaces in YCbCr space to set up complexion model.But when illumination variation is more violent, bigger variation can appear in the distribution of color in HS subspace and the foundation of CbCr subspace, this is very disadvantageous for feature detection, therefore this part utilizes the PCA conversion to set up the PCA color space, to the colour of skin sample training in the PCA color space, obtain a stable characteristics detector in conjunction with neural net.

Characteristics of image based on neural net and PCA conversion detects: the present invention proposes a kind of characteristics of image detection algorithm based on neural net and PCA conversion, this algorithm detects one by one to the pixel of input picture, under training mode, we carry out the PCA conversion to the feature samples in the training set in rgb space, obtain the projection matrix of a linearity.Secondary series vector sum the 3rd column vector of projection matrix constitutes new two dimensional character and detects the space, the axial vector that is called the PCA feature space, these two vectors are over against the direction of answering feature pixel variations minimum in rgb space, therefore, feature samples in the former training set obtains new feature samples after passing through the matrix projective transformation of being made up of secondary series vector sum the 3rd column vector, the polymerization in the PCA feature space of these feature samples is tight, at last, feature samples in the PCA feature space is delivered neural net train, until network convergence.Under detecting pattern, each pixel of image to be detected is delivered neural net after through the matrix projective transformation that is made of secondary series vector sum the 3rd column vector that obtains under training mode and is detected, and detects one by one to finish, and obtains the testing result of entire image.

The PCA feature space: following condition must be satisfied in a good feature detection space:

1. colouring information is concentrated on certain two component in the image;

2. the non-colouring information (as monochrome information) of these two components should enough lack;

3. the mean square deviation of these two components should be enough little.

The PCA conversion is the optimal mapping under the mean square error meaning, also claims the KL conversion usually.Be expressed in matrix as: A=O ^TB

In the formula, A is the vector after the conversion, and B is a vector of wanting conversion, and O is a transformation matrix, and is closely related with B, usually is made up of the characteristic vector of the autocorrelation matrix of B.So on mathematics, the core of PCA conversion is to find the solution the characteristic value and the characteristic vector of matrix.

We set up the PCA feature space by the PCA conversion.If X is the feature samples set that is used to train in the rgb space, X=[X ₁, X ₂, L, X _T], T is the number of feature samples here.The mean vector of calculated characteristics sample at first $M=\underset{i=1}{\overset{T}{\mathrm{\&Sigma;}}}{X}_i,$ It is 0 sample set that the rgb space feature samples is gone to obtain after the average average

Φ＝[Φ ₁，Φ ₂，L，Φ _T]，Φ _i＝X _i-M，1≤i≤T。Then calculate autocorrelation matrix S _T, $S_T=\underset{i=1}{\overset{T}{\mathrm{\&Sigma;}}}{\mathrm{\&Phi;}}_i{{\mathrm{\&Phi;}}_i}^T.$ Obtain autocorrelation matrix S at last _TCharacteristic value and characteristic vector, S _Tψ=ψ Λ, ψ=[ψ here ₁, ψ ₂, ψ ₃] representing the proper phasor of matrix, Λ is an eigenvalue ₁, λ ₂, λ ₃(λ ₁〉=λ ₂〉=λ ₃) diagonal matrix that constitutes.Eigenvalue ₂, λ ₃Two corresponding vectorial ψ ₂, ψ ₃Corresponding in rgb space the direction of feature pixel variations minimum, therefore with ψ ₂, ψ ₃Be considered as two main shafts in the new color space, constitute the PCA feature space, and ψ ₂, ψ ₃Constitute the linear projection matrix, the feature samples in the former rgb space arrives the PCA feature space through the linear projection matrixing.

The BP neural net: neural net method has good parallel processing performance, and good generalization ability is arranged, and does not need the prior probability distribution of data, and therefore, neural net method has embodied huge superiority in area of pattern recognition.The BP neural net is the most ripe and most widely used a kind of network of studying in the feed-forward type neural net, and we adopt the BP neural net of a hidden layer here.It is three layers that network is divided into: i is an input layer; J is a hidden node; K is the output layer node.The study error function of define grid is

$E=\frac{1}{2}\underset{k}{\mathrm{\&Sigma;}}{(d_k-y_k)}^2$

In the formula: d _kThe desired output of expression network; y _kThe actual output of expression network.So it is as follows to release each layer weights correction formula: hidden layer and output layer:

w _jk(t+1)＝w _jk(t)+ηδ _ky _j

δ _k＝y _k(1-y _k)(d _k-y _k)

Input layer and hidden layer

w _ij(t+1)＝w _ij(t)+ηδ _jy _i

${\mathrm{\&delta;}}_j=y_i(1-y_j)\underset{k}{\mathrm{\&Sigma;}}{\mathrm{\&delta;}}_k{w}_{\mathrm{jk}}$

In the following formula: η is a learning rate; δ _k, δ _jBe the corresponding correction value of each layer.

Network pornography image of the present invention and bad image detecting system, by differentiating the similar ratio of characteristic image coupling in webpage pornographic image and the pornographic standard picture feature database, the pornographic image rank can be set be tackled respectively, at adult or children, browsed content can be different.

Embodiment two: referring to Fig. 1, Fig. 3, present embodiment is substantially with embodiment one, and its difference is: system does not contain other bad image detection subsystem.Network image is after detection of process icon and text detection, isolate the webpage normal picture, suspect image is delivered the color detection subsystem to be detected, isolate the webpage normal picture, the suspect image that can't judge color detection, be sent to the attitude detection subsystem and carry out similar matching judgment, filter out pornographic image with pornographic standard picture.

Embodiment three: referring to Fig. 1, Fig. 3, present embodiment network pornography image and bad image detecting system only contain color detection subsystem and attitude detection subsystem, only those are tackled the stronger network pornography image of visual stimulus.

Claims (8)

1, a kind of content-based network pornography image and bad image detecting system, it is characterized in that: contain skin color detection subsystem and attitude detection subsystem, system sets up the Mathematical Modeling of skin color detection and attitude detection fast algorithm, the skin color detection subsystem is formed by the skin color of phase-split network image and the experiment in color of image space is compared, adopt the hsv color space to set up complexion model, the skin color of determining the people is in selected hsv color spatial distributions situation, and then computed image colour of skin degree of exposure, determine a threshold values of differentiating image colour of skin degree of exposure, distinguish normal picture and suspect image in view of the above; Described attitude detection subsystem, at first pick out the representative standard pornographic image of some, after carrying out signature analysis, extract its feature and set up the posture feature storehouse by training, it is pornographic standard picture feature database, whether as judgement is the foundation of the coupling similitude of pornographic image, by the suspect image on the network is carried out Wavelet Edge Detection, obtain an edge image, by the Wavelet Edge image is analyzed, extract marginal point, determine the boundary rectangle of object, pixel in the rectangle is cut apart according to complexion model, the skin area image of tentatively being cut apart, be converted to gray level image then through the morphologic filtering corrosion treatment, by shape description and posture analysis to the skin area image, the definition matching similarity, image in the pornographic characteristics of image of present image and the standard storehouse is mated similar judgment processing, if feature similarity in the pornographic characteristics of image of present image feature and the standard storehouse, think that so this image is a pornographic image, and tackled, otherwise adjudicating this image is normal picture.

2, network pornography image according to claim 1 and bad image detecting system, it is characterized in that: by described color detection subsystem, at first the pixel transitions with network image is the hsv color space and quantizes, be divided into L color sub-spaces, determine the total shin_count and the frequency sub_count_i of sample skin pixels in this L sub spaces of sample skin pixels then by statistical analysis, wherein satisfy i=1, Λ, L $\underset{i=1}{\overset{L}{\mathrm{\&Sigma;}}}\mathrm{sub}\_\mathrm{count}\_i=\mathrm{shin}\_\mathrm{count}$

Be distributed in the possibility of this subspace with the normalized frequency as skin pixels,

v _i＝sub_count_i/skin_count

Set the possibility threshold value T_vi of a colour of skin distribution probability, if satisfy v _i〉=T_vi, then w _i=v _iOtherwise, w _i=0; Final like this obtaining: A={A ₁, A ₂, Λ, A _L}

W＝{w ₁，w ₂，Λ，w _L}

Wherein, w _iRepresent corresponding subspace A _iDegree of membership, i.e. A _iIn color be the possibility of skin color, i=1,2, Λ L, parameter L gets 72, cluster obtains the degree of membership set W of the distribution subspace set A of skin color and A; Computed image colour of skin degree of exposure:

To arbitrary image F (x, y), x=1, Λ, M, y=1, Λ, N, (x y) is transformed into hsv color space and quantification, obtains this color of pixel subspace label, makes entire image F (x with each pixel, y) just changed into a M * N label dot matrix G (m, n), statistics G (m, n) normalization histogram Hue[k], k=1, Λ, L is by the colour of skin degree of exposure in the following formula computed image $\mathrm{Ratio}=\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}\mathrm{Hue}\left[k\right]\&times;w_k$

Utilize image colour of skin degree of exposure Ratio to distinguish normal picture and pornographic image then, take two kinds of judgement modes: (1) hard decision: determine a threshold value T_Valve, relatively Ratio and T_Valve adjudicate: if piece image satisfies Ratio 〉=T_Valve, then adjudicating this image is pornographic image; Otherwise be normal picture, the value of T_Value is taken between [0.10,0.15]; (2) soft-decision: determine a low threshold value T_Low, a high threshold T_High, relatively Ratio and these two threshold values are adjudicated: if piece image satisfies Ratio 〉=T_High, then adjudicating this image is pornographic image; If satisfy Ratio≤T_Low, then adjudicating this image is normal picture; Think under other situations that this image is a suspect image, this detector is not done judgement, and the attitude detection subsystem that passes on detects;

3, network pornography image according to claim 1 and bad image detecting system, it is characterized in that: described attitude detection subsystem, the attitude detection core algorithm mainly contains Wavelet Edge Detection, image segmentation, morphologic filtering, shape description and similarity and mates several parts:

Wavelet Edge Detection adopts the Daubechies-4 wavelet basis that the suspicious original image on the network is carried out tower wavelet decomposition, obtains LL low frequency sub-band and LH, HL, and three high-frequency sub-band of HH are utilized following formula

${E[i,j]=(E_1{[i,j]}^2+E_2{[i,j]}^2+E_3{[i,j]}^2)}^{\frac{1}{2}}$

To three types of edges synthesize an edge graph E (i, j);

Image segmentation, at first the Wavelet Edge image is analyzed, extracted four marginal points in upper and lower, left and right, and determine the boundary rectangle of object according to this, wipe then and be positioned at the outer pixel of boundary rectangle in the original color image, pixel in the rectangle is cut apart according to complexion model, to any pixel p (x, y), it is transformed into the HSV space and quantizes to obtain quantizing label k ∈ [1, Λ, L], if w _k≠ 0, then keep this pixel, otherwise, wipe this pixel, the skin area image of tentatively being cut apart;

Morphologic filtering adopts mathematical morphology that the image of tentatively cutting apart is handled, and filters out the noise pixel that does not belong to object area;

Shape description, after obtaining the area image of object, utilize the second order of image and 7 constant Hu squares that third moment can draw image:

φ ₁＝η ₂₀+η ₀₂

${\mathrm{\&phi;}}_2={({\mathrm{\&eta;}}_{20}-{\mathrm{\&eta;}}_{02})}^2+{4\mathrm{\&eta;}}_{11}^2$

φ ₃＝(η ₃₀-3η ₁₂) ²+(3η ₂₁-η ₀₃) ²

φ ₄＝(η ₃₀+η ₁₂) ²+(η ₂₁+η ₀₃) ²

φ ₅＝(η ₃₀-3η ₁₂)(η ₃₀+η ₁₂)[(η ₃₀+η ₁₂) ²-3(η ₀₃+η ₂₁) ²]

+(3η ₂₁-η ₀₃)(η ₂₁+η ₀₃)[3(η ₃₀+η ₁₂) ²-(η ₀₃+η ₂₁) ²]

φ ₆＝(η ₂₀-η ₀₂)[(η ₃₀+η ₁₂) ²-(η ₂₁+η ₀₃) ²]+4η ₁₁(η ₃₀+η ₁₂)(η ₂₁+η ₀₃)

φ ₇＝(3η ₂₁-η ₀₃)(η ₃₀+η ₁₂)[(η ₃₀+η ₁₂) ²-3(η ₀₃+η ₂₁) ²]

+(3η ₁₂-η ₃₀)(η ₂₁+η ₀₃)[3(η ₃₀+η ₁₂) ²-(η ₀₃+η ₂₁) ²]

Adopt 7 characteristic values of 18 characteristic values of second order to five rank normalization central moment of image and Hu square to describe a width of cloth and cut apart later skin area feature of image shape;

The similarity coupling adopts weighting Euclidean distance to carry out measuring similarity, and establishing weight vector is W _j, current image feature is φ _j, j=1 wherein, 2, K, 25; Feature database is characterized as φ _Ij', i=1,2, K, N, j=1,2, K, 25, wherein N representation feature Al Kut is levied number, definition similarity d _iFor

$d_i=1-{\left(\underset{j=1}{\overset{25}{\mathrm{\&Sigma;}}}{W}_j{({\mathrm{\&phi;}}_j-{\mathrm{\&phi;}}_{\mathrm{ij}}^{\&prime;})}^2\right)}^{\frac{1}{2}}$

Obtain N characteristic similarity d _iAfter, set thresholding T_shape, if characteristic similarity drops on interval [T_shape, 1], then think feature similarity in present image feature and the feature database, and the number Num of statistics similar features, if Num satisfies condition: Num＞T_num, wherein T_num is the threshold value of N feature similarity number in present image feature and the feature database, thinks that so this image is a pornographic image, otherwise adjudicating this image is normal picture.

4, according to claim 1 or 2 or 3 described network pornography image and bad image detecting systems, it is characterized in that: also contain the icon detection subsystem, according to the size of images ratio network image is differentiated, at first to the width and the height setting threshold values T-size of image, judge according to the size of network image then, filtering out less than this setting threshold values, be the too little bad network image that is generally icon one class of size, is normal picture greater than the then judgement of this setting threshold values; Secondly, judge, set the ratio threshold values T-logo of picture altitude and width according to the ratio of the height and the width of image, filter out laterally or most longitudinally be the network image of the fillet shape of advertiser web site and so on, the T-size value selects 32, the T-logo value selects 10.

5, network pornography image according to claim 4 and bad image detecting system, it is characterized in that: also contain the text detection subsystem, according to text image and the general difference of continuous-tone image on color is formed, by to the histogrammic analysis of color of image, choose suitable gray value as dividing histogrammic threshold values, H[i], i ∈ [0,255], get θ _Eg〉=200 are divided into low gray value and two zones of high gray value as thresholding with grey level histogram, utilize following formula to calculate the energy proportion in high gray value zone: $p_{\mathrm{eg}}=\underset{i={\mathrm{\&theta;}}_{\mathrm{gg}}}{\overset{255}{\mathrm{\&Sigma;}}}H\left[i\right]/\underset{i=0}{\overset{255}{\mathrm{\&Sigma;}}}H\left[i\right],$ To satisfy P _Eg〉=P _EGImage be judged as text image, according to identification requirement P _EGCan choose different values, generally choose P _EG〉=0.7; Perhaps different with the comentropy that general continuous-tone image is shown according to text image, choose certain gray value range Theta _Ep1≤ i≤θ _Ep2, calculate its histogram information entropy, select θ _Ep1=127, θ _Ep2=255, histogram is done normalized: $P\left[i\right]=H\left[i\right]/\underset{i=0}{\overset{255}{\mathrm{\&Sigma;}}}H\left[i\right],$ Compute histograms local message entropy: ${\mathrm{ep}}_l=-\underset{i={\mathrm{\&theta;}}_{\mathrm{ep}1}}{\overset{{\mathrm{\&theta;}}_{\mathrm{ep}2}}{\mathrm{\&Sigma;}}}P\left[i\right]\log P$ To satisfy ep _l〉=EP _LImage be judged as text image, require EP according to identification _LDesirable different value is generally got EP for text image _L≤ 2; Perhaps differentiate the result of text image, above-mentioned two kinds of methods are carried out fusion treatment: p according to colouring information _EgSelected threshold P _EG1And P _EG2And to ep _lSelected threshold EP _L1And EP _L2, then definition:

$\mathrm{EG}=\left\{\begin{array}{cc}0,& p_{\mathrm{eg}}<P_{\mathrm{EG}1};\\ \frac{p_{\mathrm{eg}}-P_{\mathrm{EG}1}}{P_{\mathrm{EG}2}-P_{\mathrm{EG}1}}& P_{\mathrm{EG}1}\&le;p_{\mathrm{eg}}<P_{\mathrm{EG}2};\\ 1& p_{\mathrm{eg}}\&GreaterEqual;P_{\mathrm{EG}2};\end{array}\right.$

$\mathrm{EP}=\left\{\begin{array}{cc}0,& {\mathrm{ep}}_l>{\mathrm{EP}}_{L2};\\ 1-\frac{{\mathrm{ep}}_l-{\mathrm{EP}}_{L1}}{{\mathrm{EP}}_{L2}-{\mathrm{EP}}_{L1}}& {\mathrm{EP}}_{L1}<{\mathrm{ep}}_l\&le;{\mathrm{EP}}_{L2};\\ 1& {\mathrm{ep}}_l\&le;{\mathrm{EP}}_{L1};\end{array}\right.$

Definition is based on the text image identification parameter of color:

$C_H=\frac{\mathrm{EG}+\mathrm{EP}}{2}$

C _H∈ [0,1]; Then work as C _HTo look like be text image to decision diagram during greater than a certain threshold parameter T_ch.

6, network pornography image according to claim 5 and bad image detecting system, it is characterized in that: contain other bad image detection subsystem, the feature samples of other specific bad image is carried out the PCA conversion in rgb color space, set up the PCA color space, in conjunction with neural net to the colour of skin sample training in the PCA color space, obtain a stable characteristics detector, the suspect image that obtains through icon detector and text detector by and the comparison of this property detector, detect bad network image and be input to the color detection subsystem and be determined further processing.

7, network pornography image according to claim 4 and bad image detecting system, it is characterized in that: contain other bad image detection subsystem, the feature samples of other specific bad image is carried out the PCA conversion in rgb color space, set up the PCA color space, in conjunction with neural net to the colour of skin sample training in the PCA color space, obtain a stable characteristics detector, the suspect image that obtains through icon detector and text detector by and the comparison of this property detector, detect bad network image and be input to the color detection subsystem and be determined further processing.

8, network pornography image according to claim 4 and bad image detecting system is characterized in that: by differentiating the similar ratio of characteristic image coupling in webpage pornographic image and the pornographic standard picture feature database, the pornographic image rank is set.

Images