117 results about "GrabCut" patented technology

The invention provides a method and a system for automatically extracting image foreground. The method includes conducting gray level processing on colored foreground images to be extracted, conducting filtering on the images subjected to the gray level processing, detecting remarkable foreground area of the filtered images, determining foreground range and conducting final foreground extraction according to the grabcut algorithm.

The invention discloses an interactive graph cutting method. The interactive graph cutting method is used for solving the problems that an existing interactive graph cutting method is low in operating efficiency, and the workload of a user is high. The interactive graph cutting method comprises the following steps that (1) the user inputs a graph, and a rectangular region containing a foreground target graph is selected through a rectangular frame in a frame mode; (2) the outer boundary of the rectangular region is extracted through the Canny edge detection algorithm; (3) a ternary graph is initialized through the outer boundary of the region with the foreground target graph, and a background region in the rectangular region is removed through the GrabCut algorithm, so that the foreground target graph is cut; (4) the foreground target graph is output. According to the interactive graph cutting method, the advantage that user interactivity of the GrabCut algorithm is low is reserved, and the defects of the GrabCut algorithm are overcome by making full use of boundary information of a foreground target with the help of the Canny algorithm under the condition that foreground / background colors are similar or shadows exist in the foreground target. In addition, due to the fact that the iterations of the GrabCut algorithm is reduced through the interactive graph cutting method, operating efficiency is greatly improved; meanwhile, due to the fact that the user does not need to draw the general outline of the foreground target graph, the workload of the user is reduced.

The invention discloses a free interaction-type GrabCut tongue segmentation method based on deep learning, which comprises a deep convolution neural network for extracting whole tongue features, a region of interest positioning network for initially detecting a tongue area, a deep convolution neural network for carrying out deep abstract feature extraction on the region of interest and a GrabCut algorithm for segmenting a tongue picture. The problem that the existing GrabCut algorithm relies too much on manual interaction in the case of tongue segmentation can be effectively solved, and the automatization of the GrabCut algorithm in tongue segmentation can be improved.

The present invention discloses a method for extracting image foreground based on the super pixel and fast trimap image. The method mainly comprises: (1) using the SLIC algorithm to perform super pixel preprocessing; (2) extracting super pixel features such as average color feature and coordinates of the centre-of-gravity location; (3) establishing a structure of weighted directed graph, and performing iterative energy minimization segmentation for the image obtained by segmenting the super pixel by using the GrabCut technology; (4) generating a rapid self-adaptation Trimap based on Raw Mask; (5) performing smooth process for the initial hard segmentation edge to obtain Refined Mask; and (6) extracting foreground image output. According to the present invention, the efficiency of image segmentation can be improved, and image Matting edge correction can be more convenient and efficient, which gains a high application value.

The invention discloses a human image segmenting method. The method includes the steps that firstly, multiple pictures containing standing human images are collected from an electric commerce website, regions where the human images are located are marked, and a training data set is generated; then, a convolutional neural network is initialized and trained by the training data set; in the picture test process, firstly, the regions where the human images are possibly located are marked in the test pictures through the convolutional neural network and serve as possible foreground regions, a Grabcut algorithm is initialized, and the regions with the human images are segmented through the Grabcut algorithm finally. According to the method, the regions where targets possibly exist are predicted through the convolutional neural network, the possible target regions serve as masks which are used for initializing the Grabcut algorithm, the problem that interaction is needed for the Grabcut algorithm is solved, and segmenting accuracy is improved.

The invention discloses a tongue image segmentation method based on single-target area segmentation. The method comprises the following steps: step 1) obtaining an original image: collecting the original image of a tongue; Step (2), image annotation: carrying out manual annotation on each collected tongue body original image by adopting a super-pixel improved Grabcut algorithm; (3) image production: forming an image pair by the manually labeled image and a corresponding tongue body original image; Step (4) training the neural network: training the single-target semantic segmentation neural network through the images of the training set; And (5) testing the neural network: segmenting the tongue image by using the trained single-target semantic segmentation neural network, and testing the neural network through the test set image. According to the tongue image segmentation method provided by the invention, a single-target area constraint depth network is adopted to label, segment and learn the existing image to generate the mask image, and then the mask image and the original image are synthesized to form the tongue segmentation image.

The invention discloses an image database establishing method suitable for training a deep convolution neural network, and the method comprises the following steps: collecting an original image in a pest RGB form, and recognizing and marking pest categories; performing color attenuation on the original image to obtain an RGB image and a corresponding HSV image; calculating respective color proportions according to the RGB image, calculating color similarity regions according to the HSV image, and calculating a significance image of the RGB image; thresholding the significance image to obtain a binary image, using a region appointed by the binary image as an initial region, and dividing a pest target in the initial region with a GrabCut algorithm; cutting the original image according to the divided region, zooming to a uniform size and storing and establishing a database. The problems of difficulty in sampling, fuzzy marking and non-uniform size when a crop pest database is established in the prior art are solved, and the problem of over-fitting caused by too less samples and single gestures in a training process of the deep convolution neural network is solved.

The invention discloses a pedestrian re-identification method in a zero-lap vision field. The method comprises quickly detecting a pedestrian target in a monitoring vision field, and tracking the pedestrian target; by means of the tracking result information and video frame information, utilizing a GrabCut algorithm to divide an accurate pedestrian area, and at the same time utilizing a human body dividing model to divide the human body into two parts: the upper part and the lower part; utilizing the divided upper and lower part sequence for the pedestrian to respectively construct an HSV three-channel sparse dictionary for the upper part and the lower part of the pedestrian; performing sparse reconstruction for the pedestrian after performing dividing preprocessing for the detected pedestrian in the other monitoring vision field; and calculating the reconfiguration error and performing fusion determination of the reconfiguration error so as to obtain the coupling similarity. The pedestrian re-identification method in a zero-lap vision field can fully utilize the pedestrian sequence information and can reduce the interference of the background information so as to realize accurate identification for the same pedestrian target in different vision fields of different cameras under different angles and different postures.

The present invention discloses a figure segmentation method based on a depth map. The method comprises the concrete steps: (1), collecting a skeleton graph, a color image and a depth map at the same time through adoption of a Kinect sensor, and finding out a human body bone area; (2) mapping node coordinates of the skeleton graph area into the depth map, finding out a minimum rectangular area including all the skeleton nodes, and determining the initial figure area in the depth map; (3) performing area iteration expansion of the initial figure area based on the depth information, and obtaining a final rectangular area including a complete figure object; and (4) taking the rectangular area including the complete figure object as a seed area, designing an energy function according to the depth information and the color information of the image pixels, and obtaining a final figure object segmentation result through adoption of a modulated GrabCut algorithm. The figure segmentation method based on a depth map is able to obtain a rectangular area including a whole figure object through adoption of depth information to obtain an accurate and complete figure object segmentation result.

The present invention relates to a multi-view video image segmentation method based on graph cut and view correlation, which belongs to the field of image processing. With the object to overcome the shortcomings seen in a prior method such as poor robustness, slow execution rate and unsatisfactory segmentation effect, a multi-view video image segmentation method is provided which comprises mainly two parts. An improved Grabcut algorithm is used to segment any single viewpoint image that is given. The introduced smooth item makes it easier to extract a target from the background and the measure to protect edge of the target ensures that the edge of an obtained target is still a smooth one. On the basis of the obtained target of viewpoint image, other targets of viewpoint images are extracted by view disparity matching and a morphological processing method. The algorithm adopts a segmentation method featuring stereoscopic targets matching and tracking in a super-pixel domain, improves the Slic super-pixel generation algorithm and redefines the energy function, which effectively raises the segmentation efficiency under the premise of ensuring a segmentation result with high accuracy and promises the algorithm for wide applications.

The invention discloses a container contour positioning method based on angular point detection. Images at the two sides of a container below are acquired through a pick-up head, by use of a container lockhole coarse positioning and tracking method, coarse positioning areas of upper and lower lockholes of the images are obtained and images I1 and I2 in the areas are obtained; mask matrixes m1 and m2 in the same size as the images I1 and I2 are newly established, according to the upper and lower lockhole course positioning images I1 and I2 and the corresponding mask matrixes m1 and m2, segmentation images are obtained by use of a Grabcut algorithm, then according to the segmentation images, minimum external connection rectangles of foregrounds of the images are obtained by use of an algorithm of obtaining minimum external rectangles, and by taking one summit of each rectangle as an angular point of a container contour, based on a binocular stereo visual technology, pixel coordinates of inflection points are converted into world coordinates and are sequenced to form a quadrangle, i.e., the container contour. The method can effectively solve the disadvantages of light interference, non-obvious lockhole foreground and background discrimination and the like, prevents the problem of low recognition rate under the condition of insufficient light and realizes accurate positioning of the container contour.

The invention belongs to the field of image processing and relates to an obviousness detection method based on grabcut and adaptive cluster clustering. The detection method can solve the problem of image obviousness detection and comprises the steps: firstly, using a simple linear iteration clustering (SLIC) algorithm to cut an image to obtain superpixels with different sizes; then, providing an initial obviousness image of an image constructed by an adaptive clustering algorithm based on the background portion of the image obtained through the grabcut algorithm; finally, using the grabcut algorithm to update an obvious area to obtain a final obviousness image according to the obvious portion of the initial obviousness image and by searching for the obvious area through the regions of the superpixels. The method can reduce the calculated amount without losing accuracy and has higher accuracy.

The invention discloses a ship recognition system and a ship recognition method for aerial image pickup of an unmanned plane. The ship recognition system comprises a sea surface aerial image pickup database, a sea surface template selecting module, an automatic background Trimap acquiring module and a Grabcut algorithm module based on a background model. The ship recognition method comprises the following steps of establishing a sea surface template library; establishing an image feature value; selecting templates by using a voting method; designing a method matched with the template so as to grow a high-quality background by using obtained seed points; acquiring a distance threshold value parameter by the information of the template; performing region growing; generating background Trimap according to a background mask picture; initializing a background model of a grabcut algorithm by using the Trimap; and founding out the minimum iterations by which the best effect can be achieved. By the ship recognition system and the ship recognition method, a ship can be accurately and automatically recognized; when an image with a land and a sea surface which are separated from each other is provided by a user, a result processed by the method can be rapidly obtained, and a ship recognition result is obtained.

The present invention provides a method and device for location and segmentation of an optic disc in an eyeground image, and a storage medium. The method comprises in the initial location phase of an optic disc: performing improved weighting graying of a colorful eyeground image; employing the least square method to perform ellipse fitting to estimate the radius of the optic disc; employing multiple morphological operations to eliminate the interference of artery and vein vasculars, employing histogram statistics to determine an adaptive threshold to segment an area of front 0.01% brightness in the eyeground, and taking the geometric center of the area as an estimated optic disc center; and finally determining an optic disc location rectangle frame. In the optic disc accurate location and segmentation phase, the GrabCut method is employed to segment the optic disc, and the segmented area contour employs the ellipse fitting to determine the optic disc position and size parameters so as to complete the location of the optic disc. The condition of too big radius of the optic disc in the GrabCut segmentation result is subjected to secondary segmentation to determine a new threshold and perform contour fitting to obtain a final segmentation result.

The invention provides a GrabCut texture image segmentation method based on Gabor transform. The method comprises the following steps that: transforming a two-dimensional image into frequency domain space representation through Gabor wavelet transform, calculating the parametric transform of a frequency spectrum space two-dimensional Gabor filter group to obtain multi-scale multi-direction texture features, constructing a multi-scale texture energy generic function, calculating the GMM (Gaussian Mixture Model) probability distribution model of the texture features, and introducing the GMM of the texture features into GrabCut for image segmentation. The GrabCut texture image segmentation method based on the Gabor transform has the characteristics of being simple in artificial interaction, high in processing speed, good in image segmentation effect and the like.

The invention relates to the technical field of computer vision, image processing, pattern recognition and the like and particularly relates to a multiple foreground object image interactive segmentation method. The image interactive segmentation method includes the steps of (1) constructing an image pixel similarity matrix, (2) acquiring image pixel label information; (3) constructing a spectral clustering segmentation model by combining the image pixel similarity matrix and the image pixel label information and solving to obtain a preliminary segmentation result, (4) constructing spatial smoothing constraint, and (5) constructing a markov random field model by combining the preliminary segmentation result and the spatial smoothing constraint and solving to obtain a final segmentation result. According to the multiple foreground object image interactive segmentation method, advantages of grabcut methods and linear constraint spectral clustering methods are integrated, simultaneously defects of the grabcut methods and the linear constraint spectral clustering methods are overcome, and images with randomly distributed colors and multiple foreground objects can be segmented merely by labeling an extremely small quantity of pixel points.

The invention discloses a GrabCut algorithm-based automobile segmentation method of a tongue diagnosis image. The method comprises the following steps: 1, positioning a tongue body in a thresholding manner based on regression of a hue histogram; 2, over-segmenting and initializing a central area as a foreground and a peripheral area as a background; and 3, carrying out automatic segmentation on the tongue diagnosis image obtained in the step 2 according to a GrabCut algorithm. Compared with the prior art, the automatic segmentation method disclosed by the invention is not independent of an assumption that a tongue body is positioned in the center of the image and is more effective by means of techniques of sequentially zooming out, processing and zooming up. An experiment on a true tongue diagnosis image indicates that the automatic segmentation method can be used for effectively finishing automatic segmentation of the tongue diagnosis image and making preparation for further analysis of the tongue diagnosis image.

The invention relates to a portrait photograph automatic background blurring method based on saliency detection. The method comprises the following steps: 1) segmenting a portrait image into N superpixels through a linear spectrum clustering superpixel segmentation algorithm, and calculating a saliency value of each superpixel through an improved saliency optimization algorithm; 2) marking the superpixel, the saliency value of which is greater than an adaptive threshold, as a foreground area through an Otsu method, marking the superpixel, the saliency value of which is smaller than a fixed threshold, as a background area, and marking the rest superpixels as an unknown area to obtain a superpixel scale marked three-parted graph; 3) carrying out segmentation on the marked three-parted graphto obtain a portrait area boundary by utilizing a superpixel scale GrabCut algorithm; and 4) carrying out blurring on the background area through a fast guided filtering algorithm first, and then, selectively carrying out detail enhancement on the foreground area according to the saliency detection result to obtain a background blurring effect. The method can quickly carry out background blurringby only relaying on a single portrait image, and improves the background blurring effect.

The invention provides a segmentation method combined with a C-V level set and a GrabCut algorithm. The method comprises the steps that a tooth image is preprocessed by image histogram information anda morphological method to highlight the contour of a target tooth; the contour of the target tooth is acquired through a C-V level set method, and is marked as a binary image; the binary image undergoes pyramidal up-sampling and down-sampling and is stored separately as the mask of the GrabCut algorithm; and the GrabCut algorithm is directly used to acquire the target tooth. Tooth segmentation iscarried out by combining the C-V level set with GrabCut, which effectively overcomes the segmentation difficulties of the tooth image with uneven grayscale distribution and crown adhesion. The segmentation method provided by the invention has the advantages of better segmentation effect and higher segmentation accuracy, and the average deviation degree of segmentation is 0.0198.

The invention relates to a specimen-based automatic lepidoptera insect species identification method. The method comprises the following steps: removing background by using LazySnapping and / or GrabCut, segmenting left and right wings by using a projection-based method, and determining critical points by using an edge fitting method and aligning the positions of the left and right wings; carrying out feature extraction in three RGB (Red, Green, Blue) channels, dividing a wing part into a plurality of cells in radius and angle directions, and calculating pixel average values of each cell to construct final feature vectors; and realizing the classification identification by using an SVM (Support Vector Machine) classifier. The lepidoptera insect specimen image identification method is simple and convenient to operate, high in recognition precision and strong in fault tolerance, has ideal time performance, and can be used for obviously improving the lepidoptera insect species identification efficiency.

The invention provides a method for fast segmenting a kidney CT sequential image. The method comprises the steps that (1), according to a picture including the kidney part in a given CT image sequence, a gray threshold capable of approximately distinguishing the kidney area and the perimeter area is manually determined; (2), the initial kidney rectangular region of interest (ROI) of the CT sequential image is obtained; (3), the ROI is optimized; (4), the ROI of the current sequential image is segmented through a GrabCut algorithm; (5), TLD tracking is carried out on the sequential CT image through the obtained ROI, and an updated ROI is obtained; (6), the steps 3,4 and 5 are sequentially executed in a circulating mode until the image sequence is finished, and the fast and accurate segmentation of the whole sequential image is completed. The processing area of the segmentation operation is reduced to the maximum degree, and processing speed is improved. The ROI optimizing method is provided for guaranteeing precision, and finally the reliable segmentation result is obtained.

The invention provides an automatic detection method of a salient object based on salience density and edge response, and relates to a method for automatically detecting the salient object, solving the problems of the convectional salient object detection method that only one attribute that is the salience is utilized, but the edge attribute of the salient object is not taken into account, therefore, the detection accuracy of the salient object is relatively low. The automatic detection method of the salient object based on the salience density and the edge response comprises the following steps of: calculating and generating a salient map S of an input map according to the regional salience calculation method in combination of the global color comparison and the color space distribution; generating an edge response map E on the salient map S by utilizing a group of Gabor filters; efficiently searching a global optimal sub-window containing the salient object in the input map by utilizing the maximized branch-and-bound algorithm of the salience density and the edge response; adopting the obtained optimal sub-window as the input; initializing the GrabCut graphic cutting method; carrying out the GrabCut graphic cutting method; and automatically extracting the salient object with a good edge. The automatic detection method is applicable to the image processing field.

The invention discloses a Camshift algorithm for tracking a centroid correction model on the basis of Grabcut and an LBP (Local Binary Pattern). A target object is separated from an environment through the constant value enhancement tracking of a video stream and Grabcut foreground segmentation to cause the Camshift to obtain a pure histogram. Meanwhile, a Kalman filter assists the Camshift to predict a target movement locus. During a tracking period, carrying out LBP transform on an image in a target frame to obtain a template and current LBP histogram data, a judgment coefficient and a frame body change situation are obtained through comparison, and an S-Grabcut algorithm is executed if the target object is blocked by an object with a similar color, a centroid is removed, and normal tracking is continuously carried out. Compared with a traditional Camshift algorithm, the algorithm disclosed by the invention reduces the interference of background noise to a large extent, and the problem of quick movement and blocking is solved since the Kalman filter is added. Meanwhile, interference brought by the blocking of the object with the similar color can be favorably solved by the centroid correction model. Experiment results indicate that the algorithm has good robustness, meets the requirements of instantaneity and accuracy in tracking and causes the target to be more stably tracked under a complex environment.

The invention discloses a GrabCut automatic segmentation algorithm in combination with RGBD data. Firstly, using data such as depth and color collected by a depth camera Kinect V2 for preliminarily segmenting a foreground containing a person through a coordinate system of a depth map corresponding to pixel points of a color map, and the foreground containing the person is used as a GrabCut mask frame; and then taking the depth data as a fourth channel of the Gaussian mixture model to improve an energy equation of a GrabCut algorithm, thereby realizing automatic segmentation of the figure contour. The improved Grabcut algorithm solves the problems that the segmentation is incomplete when the clothing color difference of the upper and lower bodies of the figure is large, the segmentation isinaccurate when the colors of the foreground and the background are similar and user interaction is needed in the original algorithm under a complex static background, and has the characteristics of automatic figure contour segmentation, strong real-time performance, high segmentation accuracy and the like. And drawing a figure contour by further utilizing polygon fitting to realize contour vectorization.

The invention discloses a multi-temporal satellite remote sensing island bank line and development and utilization information extraction method, and belongs to the crossing field of ocean island satellite remote sensing and computer graphic processing. The method comprises the following steps: firstly, analyzing imaging characteristics of different types of island bank lines and development and utilization conditions; secondly, adopting an image downsampling and upsampling processing technology to realize rapid extraction of a large-breadth satellite remote sensing image, and combining with adistance regularization geometric active contour model to carry out accurate step-by-step approximation to obtain an island bank line; then, a adopting a Grabcut algorithm to segment the ocean background and the island reef foreground, and utilizing color information and boundary information in the image to obtain a good segmentation result; and finally, orthogonally projecting the data to a linear subspace formed by principal components by adopting a principal component analysis method, performing feature extraction and data compression through linear transformation, and performing clustering according to the extracted features to obtain a final change detection result. The method can achieve quick and effective extraction of the island bank line and development and utilization of the high-resolution satellite remote sensing image.

The invention discloses a real-time contour refining image matting method and a storage medium, and the method comprises the steps: obtaining the background knowledge before the priori, building a buffer mask along the contour of a foreground, and roughly obtaining the foreground range; Downsampling the mask data in the buffer area to a certain proportion; In combination with buffer area mask data, segmenting the color image based on speed optimization and a Grabcut algorithm of adding figure contour shape priori to obtain a segmentation result of the scale; And upsampling the segmentation result to the same resolution as the original data to realize image matting. According to the invention, the original Grabcut algorithm is optimized; The speed of the optimized algorithm is increased by30%, the segmented contour is high in robustness, is not influenced by surrounding objects with large gradients, and moves basically along a real target contour, human-computer interaction is not needed, a character foreground can be extracted in real time under a high-resolution video flow without depending on an expensive GPU, and the fineness of the edge can be guaranteed.

The invention discloses an automatic retrogradation method and a retrogradation device in the typesetting process, and the technology is based on a GrabCut method that is applied to the typesetting area. In addition, the invention also comprises the steps that a memory block is allocated to every colorful space of the image according to the different colorful spaces of the image, and a rainbow pseudo-color coding is adopted to convert a gray image and a binary image into an image having three colors such as red, green and blue, and for the image having larger size, the weighing, the average value, the covariance and the Gaussian probability distribution within the range of a pixel point 4 are calculated. Through the technical proposal of the invention, the automatic retrogradation of the image containing different colorful spaces in the typesetting process is manually completed alternatively, and a correct partition result can be obtained; at the same time, a cancel operation can be carried out during the automatic retrogradation process, thereby being convenient for the user to complete the automatic retrogradation or re-select the retrograding area.

The invention discloses a multi-view fusion-based image foreground automatic extraction method. The invention mainly aims to solve the problems of misellaneous extraction processes and the inaccuracyof extracted foreground edges in the prior art. According to the method provided by the technical schemes of the invention, an SVM classifier is trained at first, then the grayscale image of an imageto be extracted is obtained; a sub-image containing a foreground is detected from the grayscale image through the trained SVM classifier; the position coordinates of the sub-image in the image to be extracted are used as the input of the GrabCut algorithm, and foreground extraction is performed on the image to be extracted, and the extraction result of the image to be extracted is obtained from apixel view; an image of the image to be extracted from a superpixel view is generated with the SLIC algorithm; and the image from the superpixel view and the extraction result from the pixel view arefused, so that the foreground extraction result of the image to be extracted is obtained. With the method of the invention adopted, a foreground extraction process can be simplified, and the efficiency and precision of extraction are improved. The method can be used for stereo vision, image semantic recognition, three-dimensional reconstruction and image search.

The invention discloses an underwater sea cucumber image segmentation method based on significance and Grabcut, and relates to the technical field of digital image analysis and processing. The methodis characterized by comprising the following steps: defogging a collected image by using a Retinex algorithm to increase the image contrast; using an SLIC algorithm to generate a super-pixel graph, and improving the precision of target segmentation; performing target detection on the superpixel image by using a manifold sorting significance algorithm, obtaining a multi-scale significance image ofthe target through multi-scale manifold sorting, and performing weighted fusion to obtain a final significance image of the target; and combining a GrabCut algorithm with the saliency image to realizeinteractive-free segmentation of the sea cucumber target to obtain a final sea cucumber segmentation image. By means of the method, the problem that the GrabCut algorithm depends on manual participation is solved, the segmentation algorithm time is shortened, the GrabCut segmentation efficiency is improved, and automation of GrabCut underwater sea cucumber segmentation is achieved.

The invention discloses a method for automatically identifying and segmenting a salt body through weak supervised learning. The method comprises the steps of: preprocessing a salt body picture by using a Grabcut algorithm; preliminarily segmenting a salt body contour; obtaining a training set, establishing a weak supervision segmentation model by using a convolutional neural network, utilizing a loss function to guide a model to learn features of a salt body, continuously correcting labels in a training set in an iterative mode, wherein the labels refer to salt-containing labels and salt-freelabels, the labels in the training set are more accurate pixel by pixel, and a trained weak supervision segmentation model is obtained and used for segmenting the position contour of the salt body ina to-be-tested picture. According to the method, whether the salt body exists or not and the existing position and contour can be automatically and accurately recognized by a machine, the characteristics of the salt body can be learned under the condition that data set labeling is incomplete and inaccurate, and efficient and accurate segmentation of the salt body is achieved.