CN107452010A - A kind of automatically stingy nomography and device - Google Patents

Abstract

A kind of automatically stingy drawing method and device are related to digital image processing field, including：The original image of figure to be scratched is obtained, it is calculated and scratches figure visual saliency；Foreground area, background area are isolated with filter in spatial domain and Threshold Segmentation Algorithm, combining form student movement calculates to obtain three components；Gradient calculation is carried out to each pixel of zone of ignorance, is sampled to obtain the prospect of current unknown region pixel, background sample point set according to gradient direction and conspicuousness size；The opacity and confidence level of each sample point are calculated, takes confidence level highest sample to as the final optimal sample pair for scratching figure.The regional area of smooth opacity, the opacity finally estimated；Finally, according to the opacity and the color of optimal sample pair finally estimated, carry out scratching graphic operation in original image, extract foreground target.The invention also discloses a kind of automatically stingy map device.The embodiment of the present invention have without user mutual, it is easy to use, scratch the advantages of figure precision and success rate is high.

Description

An automatic matting algorithm and device

technical field

The invention relates to the field of digital image processing, in particular to an automatic image cutout algorithm and device.

Background technique

In real life, the object of interest is extracted from a background image and used as an independent material or synthesized with a new background image in order to obtain a complete and realistic background replacement effect. This technology has been widely used in image editing, Fields such as film and television special effects have widely penetrated into people's daily lives. Due to its good application prospect and commercial value, digital image matting technology has become a hot spot in the field of computer vision research in recent years.

Digital matting algorithms model each pixel in a natural image as a linear model of foreground and background colors, namely:

I=αF+(1-α)B (1)

Among them, I represents the color value in the actual image, F represents the foreground color value, B represents the background color value, α is called the foreground opacity, and the value range is [0,1], wherein the opacity of the foreground area α=1, The opacity of the background area is α=0, and in the unknown area, that is, the edge area of the foreground object, α takes a value between (0,1). The so-called matting is the process of finding the foreground F, the background B and the opacity α when the actual image I is known. I, F, and B are all three-dimensional vectors, and the equation needs to obtain the remaining 7 unknown quantities based on 3 known quantities, so it is a highly underconstrained problem.

At present, the matting technology that has been widely used in film and television and media production companies is blue screen matting. Its principle is to limit the background to a single blue color, thereby compressing the unknown quantities in the equation into four. The blue screen cutout is easy to operate, but it has great restrictions on the background, and when the foreground appears blue, the target cannot be completely cut out.

At present, the natural image matting algorithms mainly studied by scholars can be roughly divided into two categories, namely:

(1) Sampling-based algorithm. This method assumes that the image is locally continuous, and estimates the foreground and background components of the current pixel with known sample points near the unknown area. For example, the invention CN105225245 proposes a natural image matting method based on texture distribution assumptions and regularization strategies, which improves the Bayesian matting method, but the defect of the sampling-based method is that the obtained alpha map has poor connectivity and often requires image priors Knowledge and a large number of user tokens;

(2) Algorithms based on propagation. This method requires the user to first mark (such as points, lines, etc.), identify the foreground and background, and then treat the unknown area as a field, and the edge of the field corresponds to the known area. By establishing a Laplace matrix to describe the relationship between alpha values, The disadvantage of transforming the matting process into the solution process of Laplace matrix is the large amount of calculation, and the effect is not good for non-connected regions.

In addition, there are algorithms that combine sampling and propagation in order to take advantage of the two, such as robust matting algorithms, etc., but the algorithms still generally have defects such as complex user interaction, too many prior assumptions about images, and large amounts of calculations. , thereby limiting the scope of application and increasing the difficulty of use.

Contents of the invention

In order to solve the problems existing in the prior art, the present invention provides an automatic image matting algorithm and device, which calculates the visual saliency of matting based on the input image, without limiting the background and prior knowledge of the image, from the natural scene image Fully automatic matting can be completed without user interaction, while ensuring high matting accuracy and success rate.

The technical solution adopted by the present invention to solve technical problems is as follows:

An automatic image matting algorithm, the method comprising the steps of:

Step 1: Obtain the original image to be matted, and calculate its matting visual salience;

Step 2: According to the matting visual saliency map, use spatial domain filtering and threshold segmentation algorithm to separate the foreground area and background area, and combine the morphological operation to obtain the three-part map;

Step 3: Calculate the gradient of each pixel in the unknown area according to the tripartite map, and obtain the foreground and background sample point sets of the pixels in the current unknown area by sampling according to the gradient direction and saliency;

Step 4: Calculate the opacity and confidence of each sample point according to the foreground and background sample point sets of pixels in the current unknown area, and take the sample pair with the highest confidence as the best sample pair for the final image matting. Then smooth the local area of opacity to get the final estimated opacity;

Step 5: According to the final estimated opacity and the color value of the best sample pair, a matting operation is performed in the original image to extract the foreground target.

An automatic drawing device, said device comprising:

Image acquisition module, used to collect the color value of a single image;

The matting visual salience calculation module is used to calculate the matting visual salience of the image according to the image color value acquired by the image acquisition module;

The tripartite graph calculation module is used to separate the foreground area and the background area according to the matting visual saliency map obtained by the matting visual saliency calculation module, using spatial domain filtering and threshold segmentation algorithm, and to calculate in combination with morphological operations Get the three-point map;

The sample point set acquisition module performs gradient calculation on each pixel in the unknown area according to the tripartite map acquired by the tripartite map calculation module, and obtains the foreground and background sample points of the pixels in the current unknown area by sampling according to the gradient direction and significance set;

The opacity calculation module calculates the opacity and confidence of each sample point according to the foreground and background sample point sets obtained by the sample point set acquisition module, and takes the sample pair with the highest confidence as the best sample for final matting right. The local area of opacity is then smoothed to obtain the final estimated opacity;

The foreground cutout module is used to perform a cutout operation in the original image to extract the foreground target according to the final estimated opacity and the color value of the best sample pair.

The beneficial effects of the present invention are: the method for calculating the visual salience of matting proposed by the present invention simulates the visual attention mechanism of the human eye, can automatically extract foreground objects, eliminates complicated user interaction operations, completes the automatic matting process, and is easy to operate Convenient; by limiting the number of sample point pairs, the matting time is shortened; both the saliency map and the opacity are smoothed, and the matting accuracy is improved.

Description of drawings

Fig. 1 is a schematic flow chart of an automatic map-cutting algorithm of the present invention

Fig. 2 is a schematic flow chart of calculating regional saliency in the present invention

Fig. 3 is a schematic structural view of an automatic image-cutting device of the present invention

detailed description

The present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments.

Fig. 1 is a schematic flow chart of an embodiment of an automatic map-cutting algorithm of the present invention. The embodiment of the present invention provides an automatic map-cutting method, which can be performed by any map-cutting device with image storage and display functions. It can be various terminal devices, such as personal computers, mobile phones, tablet computers, etc., or digital cameras, video cameras, etc., and can be specifically implemented by software and/or hardware. As shown in Figure 1, the method of the present embodiment includes:

Step 1: Obtain the original image to be matted, and calculate the matting visual saliency of the original image.

Considering that the foreground target that usually needs to be extracted has the following characteristics: the foreground target has a complete target area, and the contrast with the surrounding background is obvious; the color distribution is relatively uniform; most areas are brighter; there is a more obvious edge distinction from the background. Therefore, the visual matting saliency calculation method proposed by the embodiment of the present invention takes into account the characteristics of the foreground object's color, brightness, and regional integrity. Assuming that the acquired image is in rgb format, firstly, the grayscale image needs to be calculated according to the r, g, and b color channels I _gray , the specific formula is:

I _gray = (r+g+b)/3 (2)

The original image can also be in any format such as YUV. The embodiment of the present invention does not limit the output image format of the camera, and the corresponding color-to-grayscale conversion formula also needs to be adjusted.

Then low-pass filtering and downsampling are performed on I _gray , that is, the original grayscale image I _gray is used as the 0th scale layer of the pyramid. The 1st scale layer is convolved with the 0th scale layer and the low-pass filter, and then obtained by 1/2 sampling in the x and y directions, and the rest of the layers are analogous, and the resolution of each layer is half of the previous layer. . The low-pass filter here may be a Gaussian filter, a Laplacian filter, or a Gabor filter, and the embodiment of the present invention does not specifically limit the form of the low-pass filter for generating the scale pyramid.

Corresponding to the visual characteristics of the human eye, where the brightness of the image is very low, it is difficult to attract the attention of the human eye, so it is necessary to suppress the threshold value of the brightness component of the scale pyramid, that is, the area below 5% of the maximum brightness value I _{gray_max} , the brightness The component is set to 0, which can effectively suppress the interference of the dark background. The points with low local brightness on the edge of the target may also be eliminated, but on the one hand, these parts will be preserved in the regional saliency map, and on the other hand, the edge area will be divided into unknown areas, so the final matting is complete. Sex doesn't matter.

After building the scale pyramid, the brightness saliency map is then calculated. The specific calculation method is: take the image on the fine scale c as the central area of vision, and the image on the coarse scale s as the peripheral area of vision, let the central scale c∈{2,3,4}, the peripheral scale s=c+δ , δ∈{3,4}, six central-peripheral combinations can be obtained, respectively {2-5,2-6,3-6,3-7,4-7,4-8}. Interpolate the image on the scale of s to the scale of c, and according to the formula I(c, s)=|I(c)ΘI(s)(3), subtract the interpolated image from the image on the scale of c to obtain the brightness Difference map. where I(σ) is the image pyramid, σ=0,1,...,8 represent different scales, and Θ represents the central-peripheral difference operator. These feature maps represent the difference in brightness between a certain position in the image and its local neighborhood. The larger the difference, the higher the brightness significance in the local range, and the easier it is to attract the attention of the human eye. After calculating 6 groups of luminance difference maps according to this, these difference maps are fused, redundant features are discarded, and the final luminance saliency map is generated. Since the absolute value of the difference map does not reflect the significance information on different scales, it is not possible to simply add the difference maps of different scales. In the embodiment of the present invention, the difference map is normalized, and the normalization function The specific steps are:

(1) Normalize all 6 difference maps to the interval [0,1];

(2) Calculate the local variance of each difference map separately;

(3) The fusion weight is positively correlated with the local variance, that is, the larger the local variance, the greater the amount of information contained in the difference map, and the difference map should be given a larger weight in the weighted combination.

Then, according to the formula (4) Weighting and combining the luminance difference map to obtain the luminance saliency map. in Represents additive factors across scales.

Then calculate the color saliency map. According to Ewald's "color pair" model, in the center of the human visual receptive field, if a neuron is activated by R color, it will be inhibited by G color, and if it is activated by B color, it will be inhibited by Y color. Therefore, according to the formula (21～24) Convert the image from three channels of rgb to four channels of RGBY:

R=r-(g+b)/2 (5)

G=g-(r+b)/2 (6)

B=b-(g+r)/2 (7)

Y＝(r+g)/2-|r-g|/2-b (8)

Then, according to the characteristics of human visual cells, calculate the red-green and blue-yellow anti-color groups RG and BY, that is: The specific process of calculating RG is: on the scale c and scale s respectively, subtract the R and G channels of the image pixel by pixel to obtain the absolute value, then interpolate the calculation results on the s scale to the c scale, and finally compare it with the c scale The original result above is differenced pixel by pixel. The process of obtaining BY is similar. Such multiple subtractions can obtain 6 color difference significance maps on RG and BY respectively.

Then, according to the formula Color difference maps are weighted and combined to obtain color saliency maps.

Then calculate the regional saliency map. Perform superpixel segmentation on the foreground target, then count the normalized color histogram of each superpixel, and cluster the superpixels according to the color histogram, and divide the image into several regions {r ₁ ,r ₂ ,.. .,r _k }, the cluster center of each region is ce _i , i=1,...,k. Then the region saliency map VA _r of region r _i can be calculated as follows:

Among them, w(r _i ) represents the weight of the area of r _i area, and the calculation method is:

Wherein, PN(r _i ) represents the number of pixels in the region r _i . The above formula indicates that the larger the area of the region, the greater the weight, that is, the area with a large area around the area r _i has a greater impact on its significance than the area with a small area.

Dr(r _j ,r _i ) represents the distance between the cluster center of region r _j and region r _i , namely:

Among them, ce _i (m) and ce _j (n) represent the m and n color components of the color histogram ce _i and ce _j , and Dc(m, n) represents the m and n colors in the LAB The Euclidean distance of the color space.

The embodiment of the present invention proposes a region saliency extraction method combining superpixel segmentation and clustering. The specific implementation steps will be described in FIG. 2 below.

Step A) Perform superpixel segmentation on the input image. The available superpixel segmentation methods include Normalized Segmentation (NC) algorithm, Graph Cut (GS) algorithm, Quick Shift (QS) algorithm, Simple Linear Iterative Clustering (SLIC) algorithm, etc. Considering the rapidity and practicability of the algorithm, the embodiment of the present invention selects the SLIC algorithm, and the specific steps are:

1) Assuming that the total number of pixels of the image is N, pre-segment it into K*K superpixels. First, the entire image is evenly divided into K*K small blocks, and the center of each small block is taken as the initial point. Calculate the pixel gradient in the 3*3 neighborhood of each initial point, the point with the smallest gradient is the initial center O _i of the superpixel segmentation algorithm, i=0,1,...K*K-1, and assign each initial center a single label;

2) Express each pixel as a five-dimensional vector {l, a, b, x, y} of CIELAB color space and XY coordinates, and calculate the distance between each pixel and its closest center. The distance calculation formula is:

Among them, _dlab is the color difference value, _dxy is the position difference value, S is the center distance, m is the balance parameter, and Dis is the distance between pixels. For each pixel, assign the label of the closest center to it;

3) Recalculate the centers of pixels with different labels, update O _i , and calculate the difference between the old and new centers, if the difference is less than the threshold, the algorithm ends, otherwise return to step 2).

Step B: Count the normalized color histogram of each superpixel. That is, each dimension of the Lab color space is divided into several bins, and the probability of the pixel color in the superpixel falling in each bin is counted, and finally the statistical histogram is normalized.

Step C: cluster the superpixels, and divide the image into several continuous regions. Optional clustering methods include any one based on division, model, hierarchy, grid, and density clustering. In this embodiment, the density-based DBSCAN clustering algorithm is used. The specific steps are: select one of the superpixels as a seed point, use the given error threshold Eps and MinPts to search all its density-reachable superpixels and judge whether the point is a core point. If it is a core point, form a cluster area with its density-reachable points; if it is not a core point or a boundary point, select another point as a seed point and repeat the above steps; if it is not a core point but a boundary point, consider is the noise point and discarded. Repeat the above steps until all points are retrieved, and finally form several divided regions. DBSCAN can effectively deal with noise points and can divide clusters of any shape, so it is suitable for this embodiment.

Step D: Calculate the regional saliency. According to the superpixel clustering results and formulas (12)(13)(14), the regional saliency of each region is calculated.

Finally, according to the importance of color information, area information and brightness information, use the formula VA = α _g VA _g + α _c VA _c + α _r VA _r (18) and α _g + α _c + α _r = 1 (19) The three types of saliency maps are fused to obtain the final matting visual saliency map. In this embodiment, α _c =0.5, α _r =0.3, and α _g =0.2 are taken.

Step 2: According to the matting visual saliency map, use spatial domain filtering and threshold segmentation algorithm to separate the foreground area and background area, and combine the morphological operation to obtain the three-part map.

First, the spatial domain filter is used to smooth the matting visual saliency map to remove noise, and then the threshold value segmentation algorithm (such as Otsu method) is used to calculate the threshold T _va , then in the saliency map, the pixels with a saliency value greater than T _va correspond to In the foreground area, the pixels smaller than T _va correspond to the background area, so that the rough tripartite map I _tc is obtained.

Spatial domain filtering is to ensure that there are no singularities of saliency in the local area, and at the same time smooth the saliency value. Median filtering, bilateral filtering or Gaussian filtering can be selected. In order to ensure the calculation efficiency of the algorithm, in this embodiment, a Gaussian filter is selected as the smoothing filter of the matting visual saliency map, and the filter window is 3*3.

Since in the matting visual saliency map, the saliency of brightness, color saliency and regional saliency are comprehensively considered to ensure that the saliency of the foreground area is much greater than that of the background area, so there are no strict requirements for the performance of the threshold segmentation algorithm. , the threshold T _va can take a value in a wide range without affecting the result of foreground segmentation. In this embodiment, we select the Otsu threshold algorithm, and the specific steps are:

(1) Assume that the gray level of the matting visual saliency map VA is 0, 1, ..., L-1, and the total number of pixels in the image is N, and count its gray level histogram, that is, assuming that the gray level in VA is i, the number of pixels of i=0,1,...,L-1 is N _i , then the value corresponding to gray level i in the gray histogram is N _i /N;

(2) Traverse the threshold T _va from 0 to L-1, divide the pixels into two categories smaller than T _va and greater than or equal to T _va , and count the inter-class differences between these two categories:

g＝ω ₀ (μ ₀ -μ) ² +ω ₁ (μ ₁ -μ) ² (20)

Among them, ω ₀ and ω ₁ are the proportions of the number of pixels less than T _va and greater than or equal to T _va respectively, and μ ₀ and μ ₁ are the mean values of pixels less than T _va and greater than or equal to T _va respectively.

(3) After traversal, find the corresponding T _va when the inter-class difference g takes the maximum value as the final segmentation threshold.

The shape and size of the morphological operator should be selected according to the image content such as image resolution, size and shape of the foreground area. In this embodiment, the default is a circle to ensure uniformity in all directions. When performing morphological operations, in order to avoid small holes and burrs after threshold segmentation, the following morphological operations are performed on I _tc : firstly, an opening operation is performed to connect local discontinuous regions and remove holes; The corrosion operation of _e can obtain the foreground area F _g of the trimap; the expansion operation of size r _d can obtain the background part B _g of the trimap, and the area between the foreground and the background is an unknown area, so that we can obtain The subdivided _trimap It of image I.

Assuming that the gray value of the foreground area is 1 and the gray value of the background area is 0, the morphological kernel operator is used to convolve with the binary image. For the white foreground area, corrosion can reduce its boundary, while for the black background area , the expansion can shrink its boundary, and finally the unknown area is between the foreground and the background.

Step 3: Calculate the gradient of each pixel in the unknown area according to the tripartite map, and obtain the foreground and background sample point sets of the pixels in the current unknown area by sampling according to the gradient direction and saliency.

For each pixel I(x _i , y _i ) in the unknown area, calculate its gradient value Gra _i , the direction of the gradient is denoted as θ, and the calculation formula of θ is

In this embodiment, a reference foreground and background sample point pair is searched on the straight line in the gradient direction, and around the sample point pair, the size of the search area is determined according to the matting visual saliency value of the pixel at the current position. The larger the , the smaller the search range, which means that the real foreground and background points are closer to the pixel around the pixel with high salience. Then, according to the spatial distance and visual salience, five foreground and background sample pairs are searched out respectively. The specific process is as follows:

1) Let the initial value of search radius r _s be 1, count count=0;

2) On a circle with the reference foreground/background sample point as the center and r _s as the radius, calculate whether the condition is met for all pixel points p: |VA(p)-VA(p ₀ )|<T _vap , where p ₀ is the center of the search area, that is, the reference foreground or background sample point, p is the point on the search circle, and count+1 if the condition is met;

3) Determine whether count>5, if yes, stop the search; otherwise, execute r _s ++ and return to step 2).

Such a sampling strategy, on the one hand, can generate fewer sampling point pairs, reducing the complexity of subsequent matting calculations. On the other hand, sampling along the gradient direction can ensure that the foreground and background points are located in different texture regions with a high probability. , and at the same time, the sampling according to the neighborhood saliency value can ensure a certain degree of spatial and saliency similarity between the sampling points, so this sampling method can ensure a high probability of including real sampling point pairs, thereby improving the accuracy of the matting

Step 4: Calculate the opacity and confidence of each sample point according to the foreground and background sample point sets of pixels in the current unknown area, and take the sample pair with the highest confidence as the best sample pair for the final image matting. Then smooth the local area of opacity to get the final estimated opacity;

Choose a point from the foreground sample point set and the background sample point set, according to the imaging linear model, the estimated opacity is

in with Represent the color values of the mth foreground sample point and the nth background sample point, respectively. In this way, for each unknown area pixel, a total of 25 different opacity estimates are obtained, and then it is necessary to select the opacity with the highest confidence for the foreground target.

The requirements for the optimal foreground and background point pair are: 1) the linear model of formula (1) has the smallest error; 2) the foreground sample point and the background sample point have a large color difference; 3) the foreground or background sample point and The color value of the current pixel is relatively close; 4) The spatial distance between the foreground and background sample points and the current pixel is small.

According to criterion 1) and criterion 2), the linear-color difference similarity is defined as

According to criterion 3), the color similarity is defined as

According to criterion 4), the spatial distance similarity is defined as

Define the confidence function as

in, They are the linear-color difference similarity, color similarity, and spatial distance similarity. D _i is the front-background sampling radius of the unknown pixel, which is determined by the visual saliency of the image matting. The larger the visual salience, the smaller the D _i , σ ₁ , σ ₂ and σ ₃ are used to adjust the weights between different similarities. The α with the highest confidence is selected as the estimation of the opacity of the current unknown pixel, and the corresponding foreground and background sample pair is used as the best foreground and background sample pair for the final matting.

For each pixel of the unknown region, opacity is estimated point-by-point as before. In the end, it is possible that the confidence of some pixels is too low, resulting in a large error in the estimated opacity, and finally there is chromatic aberration in the cutout. Therefore, it is necessary to locally smooth the opacity of the unknown area. The factors that need to be considered during smoothing are: color value difference, spatial position difference, and saliency difference, that is, pixels with greater local color difference, farther local spatial position, and greater saliency have smaller weights. For this reason, in order to balance the impact of space domain, color value range, and salience value range, the opacity smoothing method in the present invention is as follows:

Among them, P _i and P _j represent the coordinates of points i and j, I _i and I _j represent the colors of points i and j, VA _i and VA _j represent the visual salience of points i and j in matting, σ _p , σ _c and σ _va are used to adjust the weight among the three. The opacity calculated in this way fully considers the influence of spatial position, color, and saliency, so that the closer the spatial position, the more similar the color, and the closer the salience, the closer the opacity between pixels is, which is the same as the subjective feeling of the human eye. Consistent, thus effectively eliminating the singularity of opacity and improving the accuracy of matting.

According to the four criteria of the optimal foreground-background point pair, the metric function is determined as shown in formula (30), in which the linear model conformity, the foreground-background color difference, the color difference between the target pixel and the foreground-background, and the spatial distance are comprehensively considered Four indicators. By setting the values of σ ₁ , σ ₂ , and σ ₃ , it is used to adjust the weights of different similarities. Then, according to formula (31), the opacity is smoothed to obtain the final opacity for matting. In this embodiment, the smoothing operation comprehensively considers color differences, spatial position differences and salience differences, and by changing σ _p , σ _c and σ _va , the proportions of these three in the weighting coefficients can be adjusted. For example, if attention is paid to the significance information, σ _va should be greater than σ _p and σ _c .

Step 5: According to the final estimated opacity and the color value of the best sample pair, a matting operation is performed in the original image to extract the foreground target.

The specific operation steps are: create a new image with the same size as the original image as the background, and use the calculated opacity and foreground pixel value to synthesize it with the new background according to formula (1) to obtain the final matting result.

In the example of the present invention, the matting is performed on natural images, and the specific content of foreground objects and backgrounds is not limited, as long as the foreground and the background have obvious distinguishing boundaries that can be distinguished by naked eyes.

An embodiment of the present invention provides an automatic image matting algorithm. By obtaining the original image to be matted, the matting visual saliency of the original image is calculated; then according to the matting visual saliency map, using spatial domain filtering and threshold segmentation algorithm, Separate the foreground area and the background area, and combine the morphological operations to obtain the three-part map; according to the three-part map, perform gradient calculation for each pixel in the unknown area, and obtain the foreground and background of the current unknown area pixel by sampling according to the gradient direction and significance Sample point set: Calculate the opacity and confidence of each sample point according to the foreground and background sample point sets of pixels in the current unknown area, and take the sample pair with the highest confidence as the best sample pair for the final matting. Then smooth the local area of opacity to obtain the final estimated opacity; according to the final estimated opacity and the color value of the best sample pair, a matting operation is performed in the original image to extract the foreground target. The matting visual saliency calculation method proposed by the present invention simulates the visual attention mechanism of the human eye, can automatically extract foreground objects, eliminates complicated user interaction operations, completes the automatic matting process, and is simple and convenient to operate; by limiting sample points to The amount of matting shortens the matting time; both the saliency map and the opacity are smoothed to improve the matting accuracy.

Fig. 3 is a schematic structural diagram of an automatic picture-cutting device provided by an embodiment of the present invention, the device comprising:

Image acquisition module, used to collect the color value of a single image;

The matting visual salience calculation module is used to calculate the matting visual saliency according to the image obtained by the image acquisition module;

The tripartite graph calculation module is used to separate the foreground area and the background area according to the matting visual saliency map obtained by the matting visual saliency calculation module, using spatial domain filtering and threshold segmentation algorithm, and to calculate in combination with morphological operations Get the three-point map;

The sample point set acquisition module performs gradient calculation on each pixel in the unknown area according to the tripartite map acquired by the tripartite map calculation module, and obtains the foreground and background sample points of the pixels in the current unknown area by sampling according to the gradient direction and significance set;

The opacity calculation module calculates the opacity and confidence of each sample point according to the foreground and background sample point sets obtained by the sample point set acquisition module, and takes the sample pair with the highest confidence as the best sample for final matting right. The local area of opacity is then smoothed to obtain the final estimated opacity;

The foreground cutout module is used to perform a cutout operation in the original image to extract the foreground target according to the final estimated opacity and the color value of the best sample pair.

Specifically, the matting visual salience calculation module includes:

A scale pyramid generation unit is used to perform smoothing and downsampling according to the acquired image to be cut to generate a scale pyramid;

The brightness saliency calculation unit is used to calculate the brightness saliency map according to the scale pyramid obtained by the scale pyramid generation unit, with the image on the fine scale as the central area of vision, and the image on the coarse scale as the peripheral area of vision;

The color saliency calculation unit is used to calculate the color saliency map according to the scale pyramid obtained by the scale pyramid generation unit, with the image on the fine scale as the central area of vision and the image on the coarse scale as the peripheral area of vision;

The regional saliency calculation unit is used to perform superpixel segmentation on the foreground object according to the image to be cut obtained by the image acquisition module, and cluster the superpixels according to the color histogram, and calculate the color saliency of each cluster region;

The saliency fusion unit is used to fuse the saliency image obtained by the luminance saliency calculation unit, the color saliency image obtained by the color saliency calculation unit, and the regional saliency image obtained by the area saliency calculation unit to obtain the image to be matted. Matting visual saliency map.

The tripartite graph calculation module includes:

The spatial domain filtering unit is used to select an appropriate spatial domain filtering method to smooth the matting visual saliency map;

The threshold segmentation unit is used to segment and obtain the foreground area and the background area according to the smooth matting visual saliency map obtained by the spatial domain filter unit, and obtain a rough three-part map;

Morphological operation unit: used to perform morphological operations on the rough tripartition map obtained by the threshold segmentation unit to fill holes, and obtain foreground, background and unknown regions, that is, precise tripartite maps.

The sample point set acquisition module includes:

A gradient calculation unit, configured to obtain the gradient of each unknown pixel according to the gray value of the image to be cut;

The sampling unit is used to make a straight line according to the gradient direction obtained by the gradient calculation unit, and take the first intersection point pair between the straight line and the foreground area and the background area as the initial search point. In the neighborhood of the search point, search and unknown from near to far The sample points whose pixel significance value difference is less than the threshold.

The opacity calculation module includes:

Linear-color difference similarity calculation unit: used to extract sample points pair by pair according to the sample point set obtained by the sample point set acquisition module, and calculate its linear-color similarity;

Color similarity calculation unit: used to extract sample points pair by pair according to the sample point set obtained by the sample point set acquisition module, and calculate their color similarity;

Spatial distance similarity calculation unit: used to take out sample points pair by pair according to the sample point set obtained by the sample point set acquisition module, and calculate their spatial distance similarity;

Sample screening unit: used to calculate the confidence of each pair of sample points relative to the current unknown pixel according to the similarity value obtained from the linear-color difference similarity calculation unit, color similarity calculation unit and spatial distance similarity calculation unit; select the confidence The highest opacity is used as an estimate of the opacity of the pixel at the current location.

Smoothing unit: It is used to filter the opacity obtained by the unit according to the sample, and perform local smoothing on it.

An embodiment of the present invention provides an automatic image matting device, which calculates the matting visual saliency of the original image by acquiring the original image to be matted; then uses the spatial domain filtering and threshold segmentation algorithm according to the matting visual saliency map, Separate the foreground area and the background area, and combine the morphological operations to obtain the three-part map; according to the three-part map, perform gradient calculation for each pixel in the unknown area, and obtain the foreground and background of the current unknown area pixel by sampling according to the gradient direction and significance Sample point set: Calculate the opacity and confidence of each sample point according to the foreground and background sample point sets of pixels in the current unknown area, and take the sample pair with the highest confidence as the best sample pair for the final matting. Then smooth the local area of opacity to obtain the final estimated opacity; according to the final estimated opacity and the color value of the best sample pair, a matting operation is performed in the original image to extract the foreground object. The matting visual saliency calculation method proposed by the present invention simulates the visual attention mechanism of the human eye, can automatically extract foreground objects, eliminates complicated user interaction operations, completes the automatic matting process, and is simple and convenient to operate; by limiting sample points to The amount of matting shortens the matting time; both the saliency map and the opacity are smoothed to improve the matting accuracy.

The embodiment of the present invention also provides a computer program product for automatic image cutout.

Claims (10)

1. an automatic map-cutting algorithm, is characterized in that, the method comprises the steps: Step 1: Obtain the original image to be matted, and calculate its matting visual salience; Step 2: According to the matting visual saliency map described in step 1, use spatial domain filtering and threshold segmentation algorithm to separate the foreground area and the background area, and combine the morphological operation to obtain the three-part map; Step 3: Gradient calculation is performed on each pixel in the unknown region of the tripartite image described in step 2, and the foreground and background sample point sets of the pixels in the current unknown region are obtained by sampling according to the gradient direction and saliency; Step 4: Calculate the opacity and confidence of each sample point according to the foreground and background sample point sets of pixels in the current unknown area described in step 3, and take the sample pair with the highest confidence as the best sample pair for the final image matting , and then smooth the local area of opacity to get the final estimated opacity; Step 5: According to the final estimated opacity and the color value of the best sample pair described in step 4, a matting operation is performed in the original image to extract the foreground object. 2. a kind of automatic map matting algorithm according to claim 1, is characterized in that, the concrete steps of calculating the matting visual saliency of original image are: Step (1), calculate the grayscale image I _gray , and carry out layer-by-layer smoothing and down-sampling to I _gray to generate a scale pyramid of n layers; Step (2), taking the image on the fine scale c as the central area of vision, and the image on the coarse scale s as the peripheral area of vision, first calculate the brightness difference feature map: I(c,s)=|I(c)ΘI(s)| where I(σ) is the image pyramid, σ=0,1,...,8 represent different scales, c∈{2,3,4} is the central scale, s=c+δ, δ∈{3,4 } represents the surrounding scale, Θ represents the central-peripheral difference operator, and finally six brightness difference feature maps are obtained; the brightness saliency map is the normalized weighted sum of the six brightness difference feature maps: <mrow><msub><mi>VA</mi><mi>g</mi></msub><mo>=</mo><munderover><mrow><mi></mi><mo>&amp;CirclePlus;</mo></mrow><mrow><mi>c</mi><mo>=</mo><mn>2</mn></mrow><mn>4</mn></munderover><munderover><mrow><mi></mi><mo>&amp;CirclePlus;</mo></mrow><mrow><mi>s</mi><mo>=</mo><mi>c</mi><mo>+</mo><mn>3</mn></mrow><mrow><mi>c</mi><mo>+</mo><mn>4</mn></mrow></munderover><mi>N</mi><mrow><mo>(</mo><mi>I</mi><mo>(</mo><mrow><mi>c</mi><mo>,</mo><mi>s</mi></mrow><mo>)</mo><mo>)</mo></mrow></mrow> in, represents the normalization function, Represents the cross-scale additive factor; Step (3), calculate the image on the fine scale c as the central area of vision, and the image on the coarse scale s as the peripheral area of vision, first convert the image from the rgb channel to the RGBY quaternary color channel, and then calculate the red and green channels Color difference map RG and blue-yellow channel color difference map BY: RG(c,s)=|R(c)-G(c)|Θ|G(s)-R(s)| BY(c,s)=|B(c)-Y(c)|Θ|Y(s)-B(s)| The color saliency map is the normalized weighted sum of 12 color difference feature maps: <mrow><msub><mi>VA</mi><mi>c</mi></msub><mo>=</mo><munderover><mrow><mi></mi><mo>&amp;CirclePlus;</mo></mrow><mrow><mi>c</mi><mo>=</mo><mn>2</mn></mrow><mn>4</mn></munderover><munderover><mrow><mi></mi><mo>&amp;CirclePlus;</mo></mrow><mrow><mi>s</mi><mo>=</mo><mi>c</mi><mo>+</mo><mn>3</mn></mrow><mrow><mi>c</mi><mo>+</mo><mn>4</mn></mrow></munderover><mrow><mo>(</mo><mi>N</mi><mo>(</mo><mrow><mi>R</mi><mi>G</mi><mrow><mo>(</mo><mrow><mi>c</mi><mo>,</mo><mi>s</mi></mrow><mo>)</mo></mrow></mrow><mo>)</mo><mo>+</mo><mi>N</mi><mo>(</mo><mrow><mi>B</mi><mi>Y</mi><mrow><mo>(</mo><mrow><mi>c</mi><mo>,</mo><mi>s</mi></mrow><mo>)</mo></mrow></mrow><mo>)</mo><mo>)</mo></mrow></mrow> Step (4), perform superpixel segmentation on the foreground object, then count the normalized color histogram of each superpixel, and cluster the superpixels according to the color histogram, and divide the image into several regions {r1, r2,...,rk}, the cluster center of each region is ce _i , i=1,...,k; then the regional significance value VAr of region ri can be calculated as follows: <mrow><msub><mi>VA</mi><mi>r</mi></msub><mrow><mo>(</mo><msub><mi>r</mi><mi>i</mi></msub><mo>)</mo></mrow><mo>=</mo><munder><mo>&amp;Sigma;</mo><mrow><mi>j</mi><mo>&amp;NotEqual;</mo><mi>i</mi></mrow></munder><mi>w</mi><mrow><mo>(</mo><msub><mi>r</mi><mi>j</mi></msub><mo>)</mo></mrow><msub><mi>D</mi><mi>r</mi></msub><mrow><mo>(</mo><msub><mi>r</mi><mi>j</mi></msub><mo>,</mo><msub><mi>r</mi><mi>i</mi></msub><mo>)</mo></mrow></mrow> Among them, w(r _i ) represents the weight value of the region area, and Dr(rj, ri) represents the distance between the cluster centers of region rj and region ri; Step (5), combining brightness salience, color salience and regional salience: VA＝α _g VA _g +α _c VA _c +α _r VA _r α _g +α _c +α _r =1 Among them, α _g , α _c , and α _r are weighting coefficients of different significance. 3. a kind of automatic map matting algorithm according to claim 1, is characterized in that, obtains the method for three-part graph, and concrete steps are: First, the spatial domain filter is used to smooth the matting visual saliency map to remove noise points, and then the threshold Tva is calculated by using the threshold segmentation algorithm to obtain the rough tripartite map Itc; then the following morphological operations are performed on Itc: first The opening operation is used to connect the local discontinuous areas and remove the holes; then the erosion operation with size re is performed to obtain the foreground area Fg of the trimap; the expansion operation with size rd is performed to obtain the background part Bg of the trimap, foreground The area between and the background is an unknown area, and the subdivision trimap It of the image I to be cut is obtained. 4. a kind of automatic image matting algorithm according to claim 1, is characterized in that, described sample point set acquisition method, specifically comprises: For each pixel I(xi,yi) in the unknown area, calculate its gradient value Grai, the direction of the gradient is denoted as θ, and the calculation formula of θ is <mrow><mi>&amp;theta;</mi><mo>=</mo><mi>a</mi><mi>r</mi><mi>c</mi><mi>t</mi><mi>a</mi><mi>n</mi><mfrac><mrow><mi>I</mi><mrow><mo>(</mo><msub><mi>x</mi><mi>i</mi></msub><mo>,</mo><msub><mi>y</mi><mi>i</mi></msub><mo>-</mo><mn>1</mn><mo>)</mo></mrow><mo>-</mo><mi>I</mi><mrow><mo>(</mo><msub><mi>x</mi><mi>i</mi></msub><mo>,</mo><msub><mi>y</mi><mi>i</mi></msub><mo>+</mo><mn>1</mn><mo>)</mo></mrow></mrow><mrow><mi>I</mi><mrow><mo>(</mo><msub><mi>x</mi><mi>i</mi></msub><mo>-</mo><mn>1</mn><mo>,</mo><msub><mi>y</mi><mi>i</mi></msub><mo>)</mo></mrow><mo>-</mo><mi>I</mi><mrow><mo>(</mo><msub><mi>x</mi><mi>i</mi></msub><mo>+</mo><mn>1</mn><mo>,</mo><msub><mi>y</mi><mi>i</mi></msub><mo>)</mo></mrow></mrow></mfrac></mrow> Make a straight line along the θ direction, and get the first intersection point of the straight line with the foreground area and the background area, respectively, as the initial search center. In the neighborhood of the intersection point, search for 5 points whose significance value difference is smaller than the threshold Tvap from near to far, and finally generate a total of 5*5=25 sample point pairs. 5. A kind of automatic cutout algorithm according to claim 1, is characterized in that, opacity calculation method, specifically comprises: Choose a point from the foreground sample point set and the background sample point set, according to the imaging linear model, the estimated opacity is (Bi(t) in the formula is changed to ) <mrow><msubsup><mi>&amp;alpha;</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>=</mo><mfrac><mrow><msup><mrow><mo>(</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>)</mo></mrow><mi>T</mi></msup><mrow><mo>(</mo><msubsup><mi>Fg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>-</mo><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>)</mo></mrow></mrow><mrow><mo>|</mo><msubsup><mi>Fg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>-</mo><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>|</mo></mrow></mfrac></mrow> in with respectively represent the color values of the mth foreground sample point and the nth background sample point; thus, for each unknown region pixel i, a total of 25 different opacity estimates are obtained, and then the opacity with the highest confidence needs to be selected for use To cut out the foreground target. Define the linear-color difference similarity as <mrow><msubsup><mi>lc</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>=</mo><mo>-</mo><msub><mi>&amp;sigma;</mi><mn>1</mn></msub><mfrac><mrow><mo>|</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo><msubsup><mi>&amp;alpha;</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><msubsup><mi>Fg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>-</mo><mrow><mo>(</mo><mn>1</mn><mo>-</mo><msubsup><mi>&amp;alpha;</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>)</mo></mrow><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>|</mo></mrow><mrow><mo>|</mo><msubsup><mi>Fg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>-</mo><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>|</mo></mrow></mfrac></mrow> Define the color similarity as <mrow><msubsup><mi>co</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>=</mo><mi>exp</mi><mo>{</mo><mo>-</mo><mfrac><mrow><msubsup><mi>&amp;sigma;</mi><mn>2</mn><mn>2</mn></msubsup><mo>&amp;CenterDot;</mo><munder><mrow><mi>m</mi><mi>a</mi><mi>x</mi></mrow><mi>j</mi></munder><mrow><mo>(</mo><mo>|</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo>mo><msubsup><mi>Fg</mi><mi>j</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup><mo>|</mo><mo>)</mo></mrow><mo>&amp;CenterDot;</mo><munder><mrow><mi>m</mi><mi>a</mi><mi>x</mi></mrow><mi>j</mi></munder><mrow><mo>(</mo><mo>|</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo><msubsup><mi>Bg</mi><mi>j</mi><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>|</mo><mo>)</mo></mrow></mrow><mrow><mo>|</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo><msubsup><mi>Fg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo>< /m row></msubsup><mo>|</mo><mo>&amp;CenterDot;</mo><mo>|</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>|</mo></mrow></mfrac><mo>}</mo></mrow> Define the spatial distance similarity as <mrow><msubsup><mi>ds</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>=</mo><mi>exp</mi><mo>{</mo><mo>-</mo><mfrac><mrow><msubsup><mi>&amp;sigma;</mi><mn>3</mn><mn>2</mn></msubsup><mo>|</mo><msub><mi>x</mi><mi>i</mi></msub><mo>-</mo><msub><mi>x</mi><mrow><msubsup><mi>Fg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup></mrow></msub><mo>|</mo><mo>&amp;CenterDot;</mo><mo>|</mo><msub><mi>x</mi><mi>i</mi></msub><mo>-</mo><msub><mi>x</mi><mrow><msubsup><mi>Bg</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>)</mo></mrow></msubsup></mrow></msub><mo>|</mo></mrow><msub><mi>D</mi><mi>i</mi></msub></mfrac><mo>}</mo></mrow> 2 Define the confidence function as <mrow><msubsup><mi>c</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><mo>=</mo><mi>exp</mi><mo>{</mo><mo>-</mo><mfrac><mrow><msubsup><mi>lc</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><msubsup><mi>co</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup><msubsup><mi>ds</mi><mi>i</mi><mrow><mo>(</mo><mi>m</mi><mo>,</mo><mi>n</mi><mo>)</mo></mrow></msubsup></mrow><mrow><msup><msub><mi>&amp;sigma;</mi><mn>4</mn></msub><mn>2</mn></msup></mrow></mfrac><mo>}</mo></mrow> in They are linear-color difference similarity, color similarity, and spatial distance similarity. Di is the foreground and background sampling radius of unknown pixel i, which is determined by its matting visual salience. The larger the visual salience, the smaller Di is. σ1, σ2 and σ3 are used to adjust the weights between different similarities; select the α with the highest confidence as the estimate of the opacity of the current unknown pixel, and the corresponding foreground and background sample pairs are used as the foreground and background sample pairs for the final matting ; Finally, smooth the opacity; <mrow><msub><mi>&amp;omega;</mi><mrow><mi>i</mi><mi>j</mi></mrow></msub><mo>=</mo><mi>exp</mi><mo>{</mo><mo>-</mo><mfrac><mrow><mo>|</mo><msub><mi>P</mi><mi>i</mi></msub><mo>-</mo><msub><mi>P</mi><mi>j</mi></msub><mo>|</msub>mo></mrow><mrow><mn>2</mn><msubsup><mi>&amp;sigma;</mi><mi>p</mi><mn>2</mn></msubsup></mrow></mfrac><mo>-</mo><mfrac><mrow><mo>|</mo><msub><mi>I</mi><mi>i</mi></msub><mo>-</mo><msub><mi>I</mi><mi>j</mi></msub><mo>|</mo></mrow><mrow><mn>2</mn><msubsup><mi>&amp;sigma;</mi><mi>c</mi><mn>2</mn></msubsup></mrow></mfrac><mo>-</mo><mfrac><mrow><mo>|</mo><msub><mi>VA</mi><mi>i</mi></msub><mo>-</mo><msub><mi>VA</mi><mi>j</mi></msub><mo>|</mo></mrow><mrow><mn>2</mn><msubsup><mi>&amp;sigma;</mi><mrow><mi>v</mi><mi>a</mi></mrow><mn>2</mn></msubsup></mrow></mfrac><mo>}</mo></mrow> Among them, P _i and P _j represent the coordinates of points i and j, I _i and I _j represent the colors of points i and j, VA _i and VA _j represent the visual salience of points i and j in matting, σp, σc and σva are used to adjust the weight among the three. 6. An automatic drawing device, characterized in that the device comprises: Image acquisition module, used to collect the color value of a single image; The matting visual salience calculation module is used to calculate the matting visual salience of the image according to the image color value acquired by the image acquisition module; The tripartite graph calculation module is used to separate the foreground area and the background area according to the matting visual saliency map obtained by the matting visual saliency calculation module, using spatial domain filtering and threshold segmentation algorithm, and to calculate in combination with morphological operations Get the three-point map; The sample point set acquisition module performs gradient calculation on each pixel in the unknown area according to the tripartite map acquired by the tripartite map calculation module, and obtains the foreground and background sample points of the pixels in the current unknown area by sampling according to the gradient direction and significance set; The opacity calculation module calculates the opacity and confidence of each sample point according to the foreground and background sample point sets obtained by the sample point set acquisition module, and takes the sample pair with the highest confidence as the best sample for final matting Yes; then smooth the local area of opacity to get the final estimated opacity; The foreground cutout module is used to perform a cutout operation in the original image to extract the foreground target according to the final estimated opacity and the color value of the best sample pair. 7. A kind of automatic drawing device according to claim 6, is characterized in that, described drawing drawing visual salience calculation module comprises: A scale pyramid generation unit is used to perform smoothing and downsampling according to the acquired image to be cut to generate a scale pyramid; The brightness saliency calculation unit is used to calculate the brightness saliency map according to the scale pyramid obtained by the scale pyramid generation unit, with the image on the fine scale as the central area of vision, and the image on the coarse scale as the peripheral area of vision; The color saliency calculation unit is used to calculate the color saliency map according to the scale pyramid obtained by the scale pyramid generation unit, with the image on the fine scale as the central area of vision and the image on the coarse scale as the peripheral area of vision; The regional saliency calculation unit is used to perform superpixel segmentation on the foreground object according to the image to be cut obtained by the image acquisition module, and cluster the superpixels according to the color histogram, and calculate the color saliency of each cluster region; The saliency fusion unit is used to fuse the saliency image obtained by the luminance saliency calculation unit, the color saliency image obtained by the color saliency calculation unit, and the regional saliency image obtained by the area saliency calculation unit to obtain the image to be matted. Matting visual saliency map. 8. A kind of automatic image matting device according to claim 6, characterized in that, the three-part graph calculation module comprises: The spatial domain filtering unit is used to select an appropriate spatial domain filtering method to smooth the matting visual saliency map; The threshold segmentation unit is used to segment and obtain the foreground area and the background area according to the smooth matting visual saliency map obtained by the spatial domain filter unit, and obtain a rough three-part map; Morphological operation unit: it is used to perform morphological operations on the rough tripartition map obtained by the threshold segmentation unit to fill holes, and obtain foreground, background and unknown regions, that is, precise tripartite maps. 9. A kind of automatic image matting device according to claim 6, characterized in that, the sample point set acquisition module comprises: A gradient calculation unit, configured to obtain the gradient of each unknown pixel according to the gray value of the image to be cut; The sampling unit is used to make a straight line according to the gradient direction obtained by the gradient calculation module, and take the first intersection point pair between the straight line and the foreground area and the background area as the initial search point. In the neighborhood of the search point, search from near to far and unknown The sample points whose pixel significance value difference is less than the threshold. 10. A kind of automatic drawing device according to claim 6, is characterized in that, described opacity calculation module comprises: Linear-color difference similarity calculation unit: used to extract sample points pair by pair according to the sample point set obtained by the sample point set acquisition module, and calculate its linear-color similarity; Color similarity calculation unit: used to extract sample points pair by pair according to the sample point set obtained by the sample point set acquisition module, and calculate their color similarity; Spatial distance similarity calculation unit: used to take out sample points pair by pair according to the sample point set obtained by the sample point set acquisition module, and calculate their spatial distance similarity; Sample screening unit: used to calculate the confidence of each pair of sample points relative to the current unknown pixel according to the similarity value obtained from the linear-color difference similarity calculation unit, color similarity calculation unit and spatial distance similarity calculation unit; select the confidence The opacity with the highest degree is used as an estimate of the opacity of the pixel at the current location; Smoothing unit: It is used to filter the opacity acquired by the unit according to the sample, and perform local smoothing on it. Factors considered during smoothing include: color value difference, spatial location difference, and significant difference.