76results about How to "Splitting speed is fast" patented technology

The invention discloses a stomach computed tomography (CT) sequence image segmentation method based on interactive region growth, which mainly solves the problems that in the prior art, CT sequence segmentation speed is slow, and poor segmentation is easy to occur. The method includes: firstly, a seed point is selected manually in a target area to be segmented in a first image, the interactive region growth is used for performing segmentation, a center of a segmentation result and eight neighborhoods of the center are projected into a next CT image to serve as seed points, the interactive region growth is continuously used for performing segmentation to obtain the target area of the current image, and the segmentation result of the previous image is projected into a next image repeatedly to serve as a seed point to be segmented continuously until segmentation of a whole sequence is completed. Compared with a traditional serial region growth, the stomach CT sequence image segmentation method based on the interactive region growth has the advantages of being rapid in speed, good in effect and the like, can be used for segmenting stomach CT sequence images, and can well segment target areas which may occur in stomach lymph gland in the sequence.

The invention discloses a method for segmenting a three-dimensional ultrasonic image, belonging to the technical field of the digital image processing. The method comprises the following steps: (1) preprocessing spots of the three-dimensional ultrasonic image by adopting a normalized anisotropic diffusion method of a three-dimensional wavelet according to the characteristics of the three-dimensional ultrasonic image to remove spot noise; (2) initializing the preprocessed three-dimensional ultrasonic image by adopting a Canny edge detection operator; and (3) segmenting the three-dimensional ultrasonic image three-dimensionally by using a B-Surface and GVF Snake based three-dimensional deformation model. The method disclosed by the invention can be used to rapidly and accurately segment the three-dimensional ultrasonic image and particularly has strong noise robustness. The method for automatically segmenting the three-dimensional ultrasonic image can be also used for segmenting other three-dimensional images such as CT (computed tomography) images, MRI (magnetic resonance images) and PET (position-emission tomography) images, thereby having high application value.

The invention relates to an improved region growing method which is applied to vessel segmentation and extraction in a coronary artery angiography image. The improved region growing method comprises the following steps of: preprocessing the image to obtain an original image capable of directly performing region growth; making a regulation and randomly generating a group of seed points; setting a stack data structure, enabling a newly grown pixel point to enter a stack, and taking out the point previously entering the stack to serve as a current point to be subjected to growth when the current point completes the growth; sequentially performing growth on each seed point, wherein a seed point gray value serves as an average value at a growing initial stage, and calculating a new average gray value when a new pixel point is grown every time along with the growth of the seed points; and completing the growth when no pixel point meeting growth standards exists and no seed point exists. The improved region growing method has the advantages that the seed points are automatically generated, no manual intervention is needed, the local average values around each pixel point serve as growth parameters in a growing process, the coronary artery angiography image with uneven brightness can be segmented, and the efficiency and the accuracy of the image segmentation are improved.

The method includes following procedures. Anisotropy diffusion filtering is adopted to remove noise. Excessive segmentation is carried out for images by using Watershed method. Stack data structure is built to locate mesh point with minimum time T in narrow band. Fast marching method makes final segmentation for images. The invention raises speed of segmenting medical image greatly by using Watershed method and improved Fast Marching method, possessing wide adaptability no mater CT image or MR image. Thus, the invention has important application value in area of computer-aided diagnosis and treatment.

The invention provides a brain part MRI image segmentation method. The brain part MRI image segmentation method is characterized in that a gray level image of a brain part MRI image to be segmented can be acquired; the gray values of different pixel points of the brain part MRI image can be used as the clustering centers, which are used to form the clustering center sets as the particles, and the optimization of the clustering center sets can be carried out by adopting the particle swarm optimization algorithm; every pixel point of the brain part MRI image belongs to the category having the maximum membership, and then the gray values of the pixel points of the same category are equal to the same gray value, and the brain part MRI image segmentation can be completed. The brain part MRI image segmentation method is advantageous in that according to the chaotic characteristic and the logic self-mapping function, the uniformly-distributed particle swarms can be initialized by adopting the logic self-mapping function, and then the quality of the initial solution, the stability of the PSO algorithm, the speed and the precision of the image segmentation can be improved; the chaotic searching can be carried out, when the particles are in the premature convergence state, and the premature convergence phenomenon caused by the stagnated state of the particles during the iteration process can be prevented, and the optimal solution in the range of the whole situation can be realized, and then the speed and the precision of the image segmentation can be improved.

The invention discloses a laser radar three-dimensional point cloud segmentation method. The method comprises the following steps of firstly, extracting original three-dimensional point cloud data collected by a laser radar; preprocessing the original point cloud with steps of denoising, simplifying and coordinate transformation of original point cloud data; constructing basic point cloud data under a three-dimensional Cartesian coordinate system; storing the three-dimensional data in the form of a two-dimensional array; adopting a variable neighborhood decentralized search strategy to dynamically adjust the resolution of the neighborhood range and the search matching range of the seeds surrounded by the region growing method. Point cloud preliminary segmentation work is carried out, on the basis, a point cloud segmentation envelope diffusion strategy is designed, the periphery of a point cloud segmentation set is further searched, fusion of multiple sets is achieved, then the point cloud segmentation set is obtained, and finally a visualization function of a point cloud segmentation result is designed and used for checking the point cloud segmentation effect. The method effectively improves the segmentation rate, effectively suppresses the over-segmentation condition, maintains the integrity of each target, and facilitates the observation of the scanning result of the segmented target.

The invention discloses a computer-assisted animation image-text continuity semi-automatic generating method, and aims at providing a method for assisting animation making personnel to rapidly and efficiently make an animation image-text continuity from an original drama so as to realize standardized and text continuity automatic analysis. The technical scheme comprises the steps of: firstly, processing an original literature drama to form a standardized drama; secondly, extracting all kinds of key semantic information including story lines in the standardized drama, and editing to obtain a text continuity; thirdly, segmenting a dialog image drawn according to the standardized drama to obtain a single dialog with a scene number; and finally, matching the text continuity with the dialog according to the scene number, and guiding into a scene making introduction. By adopting the invention, the text continuity, the dialog and the making introduction of each scene can be automatically matched, working efficiency of generating the text continuity is improved, difficulty of complete dependence on manual making during the text continuity generation is solved, and speed of segmenting the dialog is greatly increased.

The invention provides a method and device for segmenting an image. The method and device for segmenting the image solves the problem that an image segmenting method in the prior art is low in efficiency. The method comprises a step of downsampling an input image to obtain a preprocessed image, a step of establishing an image date model according to the preprocessed image, a step of determining the edge outline of a target image based on the calculation of the image data model, a step of determining the position where the target image is located in the input image according to the position where the edge outline is located in the preprocessed image, and a step of extracting the target image according to the position where the target image is located in the input image. By the adoption of the technical scheme, the calculated amount in the process of segmenting can be reduced, the speed of segmenting the image can be increased, and the segmenting time can be effectively saved.

The invention discloses a hyperspectral remote sensing image vector C-V model segmentation method based on wave band selection. Firstly, according to a spectral curve, wave bands high in contrast ratio between a target and the background are selected, further, according to relevant coefficients of the wave bands, the wave bands high in relevancy are removed so that a new wave band combination can be formed, and therefore according to the determined wave band assembly, a hyperspectral image vector matrix is established; on the basis, a vector C-V segmentation model based on the vector matrix is constructed, the edge guiding function based on gradient is introduced into the model, on the basis that a traditional C-V model is reserved to perform image segmentation based on area information, the capacity for capturing the target boundary in heterogeneous areas and under complex background conditions is enhanced through edge detail information of images, segmentation precision of the hyperspectral remote sensing images is improved, segmentation speed of the hyperspectral remote sensing images is increased.

The invention discloses an MSPCNN-based gallstone ultrasonic image full-automatic segmentation method, and the method comprises the steps: carrying out the segmentation of an ultrasonic image throughemploying an MSPCNN algorithm, and obtaining a gallstone coarse segmentation binary image; Segmenting the gall bladder coarse segmentation binary image by adopting a morphological algorithm to obtaina gall bladder accurate segmentation binary image and a calculus accurate segmentation binary image; And carrying out post-processing on the accurate gall bladder segmentation binary image and the accurate calculus segmentation binary image by adopting a local weighted linear regression algorithm, enabling the edge contour of the gall bladder calculus to be smooth, and finally obtaining a gall bladder region segmentation result and a calculus region segmentation result. The advantages of reducing the calculation complexity, reducing the segmentation steps and improving the image segmentation speed and precision are achieved.

The invention provides a cotton foreign fiber image online dividing method and a cotton foreign fiber image online dividing system. The cotton foreign fiber image online dividing method includes the following steps: S1, receiving cotton foreign fiber images, converting the cotton foreign fiber images to gray-level images and negating the gray-level images; S2, processing the negated gray-level images in blocks and judging whether to further process the gray-level images in refining mode; S3, reading image blocks needing refining processing, obtaining revised background gray-level images, leading the background gray-level images to subtract the original gray-level image blocks to obtain image blocks with backgrounds removed, and cutting the images in an OTSU method; and S4, performing expansive operation on the image blocks, comparing corresponding edge pixels of adjacent image blocks, and performing image linking on cracked foreign fiber target images in multiple images or image blocks based on overlap ratio of the corresponding edge pixels. The cotton foreign fiber image online dividing method and the cotton foreign fiber image online dividing system can effectively improve imagecutting speed and ensure cutting quality of the images.

The invention discloses an image segmentation method based on an LBP operator and a chain code technology, and the method comprises the following steps: 1, carrying out the preprocessing of an image,and obtaining an image with a texture feature; 2, obtaining an LBP characteristic value through a rotation invariant LBP operator; 3, selecting the value with the minimum LBP characteristic value as an LBP operator in a rotating state, and processing the image according to the LBP operator to obtain an effect picture; 4, carrying out noise point filtering on the effect image in the step 3; and step 5, determining a segmentation region through the Freeman chain code. The problem of edge discontinuity is improved through extraction and classification of a plurality of features in cooperation with chain code detection.

The existed method has problems of: (1) difficult to divide the uneven illuminated picture with a unified threshold; (2) sensitive to the unevenness of noise and gray scale. This invention includes steps of: (1) building a oscillating net; (2) searching the oscillation starting point in the net; (3) starting the oscillation from the starting point and iterative extending to the neighbor domain; (4) after oscillation, distinguishing ' target' from 'background' depending on if possessing a mark. In picture dividing methods, This method combines the advantages of both the margin tracing method and the domain generating method (DG). It obtains the same result as DG, but greatly reduces the time / space complexity and greatly raises the toughness and the dividing speed. It will have an extensive develop foreground in practical application.

The invention provides a semi-supervised video target segmentation method, which comprises the steps: S1, preprocessing a video image to obtain an image of a current frame and an image of a first frame, and giving a segmentation image of the first frame; S2, constructing a semi-supervised video target segmentation network model, wherein the semi-supervised video target segmentation network model comprises a short-time network module, a long-time network module, an attention gate network module and an up-sampling module; S3, inputting the image of the previous frame, the segmentation result image of the previous frame and the image of the current frame into a short-time network module to obtain a rough segmentation image and relative change information of the current frame; inputting the image of the current frame, the image of the first frame, the segmentation map of the first frame and the rough segmentation map of the current frame into a long-term network module to obtain absolute change information; inputting the relative change information and the absolute change information into an attention gate network to obtain a segmentation result, and finally obtaining a segmentation result graph through an up-sampling module. According to the method, the segmentation performance and the segmentation speed can be improved.

The embodiment of the invention discloses a lung nodule image segmentation method, which relates to the fields of computer technology and medical image analysis. The lung nodule image segmentation method comprises the following steps: all chest CT images are annotated to obtain annotated CT image data set; the convolution neural network model of pulmonary nodule detection is constructed and the CTimage data set is input into the convolution neural network model of pulmonary nodule detection; super parameters of the convolution neural network model are set, the convolution neural network modelis trained by Caffe to detect the pulmonary nodules, and a training model is generated; a CT image data set is input into the training model, and the detected pulmonary nodule position information isoutput after completing the training; threshold method is used to binarize the detected pulmonary nodule region, and the main region of pulmonary nodule is obtained. The seed points are randomly selected from the main areas of pulmonary nodules and the nodules are segmented by a region growing method. The invention can solve the problem that the lung nodules can not be accurately and automatically segmented in the prior medical diagnosis technology, and the treatment is difficult.

The invention discloses an image processing method and device and a storage medium. The method comprises the steps of extracting a target area image in an original pathological image; performing imagesegmentation on the target area image to obtain a focus area image, the focus area image comprising a focus area of the target area image; and marking a focus boundary of the target area image in thetarget area image or the original pathological image according to the focus area image. According to the image processing method, after the target area image is extracted from the original pathological image, accurate focus boundary detection can be conducted on the target area image, and therefore the image processing method used for achieving area-level accurate focus boundary detection is provided.

The invention relates to the field of bone mineral density measurement, in particular to a bone mineral density measurement method and system, a storage medium and electronic equipment. The method comprises the following steps: step 1, carrying out spine area detection on a CT chest plain scanning sequence image through an optimized target detection algorithm to obtain a spine area image; step 2, inputting the spine region image into a spine segmentation model for segmentation to obtain a segmentation result; step 3, processing the segmentation result to obtain a cancellous bone region; and step 4, training a regression bone mineral density detection model based on the cancellous bone area, and inputting an image to be measured into the regression bone mineral density detection model to obtain a bone mineral density value. The method can achieve the effects of higher efficiency and higher accuracy.

The invention discloses an image segmentation method based on a convolutional network. The image segmentation method based on a convolutional network comprises the steps of 1, preprocessing data; 2, designing a convolutional network model, wherein the convolutional network is called as an LBNet network and is mainly improved based on an ENet network; 3, carrying out model training and verification; 4, performing model optimization and improvement processing, and continuously adjusting hyper-parameters of the model according to a measurement result on the test set in the step 3 to realize parameter optimization of the convolutional network model established in the step 2; and 5, model use: carrying out test use according to the finally optimized model obtained in the step 4. The image segmentation method and the image segmentation process based on the convolutional network have the beneficial effects that the convolutional network is formed by improving an ENet network as a backbone network, and an original ENet network structure is modified in the implementation process.

The invention discloses a three-dimensional model reconstructing method for keeping a fracture line of a jaw bone. The method is characterized by comprising the following steps that 1, CT data, in accordance with a Dicom protocol, of an injured jaw bone are input; 2, a surface model of the jaw bone is extracted by using a Marching Cube isosurface algorithm; 3, a segmentation algorithm which combines a Gaussian mixture model with Graph Cut is used for rapidly completing segmentation of broken bone blocks and a main bone block; 4, automatic replacement is conducted on the broken bone blocks on the basis of the symmetry, and a complete jaw bone model is obtained by splicing the broken bone blocks and the main bone block; 5, positioning devices which are each composed of a pair of coupling block devices are additionally arranged on the seams of the complete model obtained through splicing; 6, a complete three-dimensional solid model is obtained through assembling according to the positioning devices. By means of the three-dimensional model reconstructing method for keeping the fracture line of the jaw bone, in a jaw bone repairing operation, particularly when chimerism, fixation and repair are conducted between the broken bone blocks and between the broken bone blocks and the main bone block, the fracture line information of the defect parts is kept, so that errors in the operation are reduced, the wounds of a patient are reduced, the process of the operation is accelerated, and the operation effect is guaranteed.

The invention discloses a point cloud segmentation method for three-dimensional measurement of a complex special-shaped curved surface robot, and the method comprises the following steps: S100, inputting a blade point cloud X taking the ground and a desktop as backgrounds, filtering background points through voxel filtering, and obtaining a target blade point cloud Y; S200, calculating a normal vector and a plane profile tolerance of a Y midpoint by utilizing a PCA algorithm, removing outliers, and marking an associated point set as a consistent set CS; s300, establishing paired connection byutilizing the normal vector and the plane profile tolerance deviation, searching after determining a clustering center, and searching all points connected with the clustering center to generate a cluster C; s400, performing curved surface fitting on the cluster C by using a Delaunay triangulation method; s500, for each fitted curved surface slice, calculating the curvature of the curved surface slice, setting a curvature deviation threshold value, and if the curvature deviation between two adjacent curved surface slices is smaller than the threshold value, combining the curved surface slices;otherwise, not combining to obtain a complete leaf point cloud Y separated from the background point cloud. The method has the advantages of being accurate in segmentation, few in input parameters andhigh in robustness.

The application provides a method and a device for word segmentation. The method comprises the following steps: acquiring a sample image, wherein the sample image includes word interval markers or non-word interval markers; processing the sample image with a convolution neural network to get a first characteristic vector corresponding to the sample image and a word interval probability value and / or a non-word interval probability value corresponding to the first characteristic vector; acquiring a to-be-tested image, and processing the to-be-tested image with the convolution neural network to get a second characteristic vector corresponding to the to-be-tested image and a word interval probability value and / or a non-word interval probability value corresponding to the second characteristic vector; and carrying out word segmentation on the to-be-tested image according to the word interval probability value or the non-word interval probability value obtained currently. Through the technical scheme of the application, word segmentation is carried out accurately. Therefore, the accuracy of word segmentation is improved, the speed of word segmentation is improved, and the user experience is improved.

The invention discloses an eye fundus image blood vessel segmentation method and system based on self-supervised learning. The method comprises the following steps: improving a U-net structure by a used network model, mutually transmitting different layers of feature maps to meet requirements of eye fundus image detail feature extraction, and increasing the speed of eye fundus image blood vessel segmentation in a segmentation process through network pruning; then designing and adopting an aggregation task strategy, and combining four methods of intensity transformation, random pixel filling, inward filling and outward filling so as to obtain more global features and detail features of an eye fundus image in a pre-training learning process; finally, designing a vector classification task module to generate different vector routes, and training an encoder through a network prediction vector route to obtain spatial correlation features of the eye fundus image. According to the method, effective eye fundus image features can be learned from unlabeled data, and blood vessel segmentation precision equivalent to that of a supervised deep learning method can be achieved with fewer training iterations and manual labeled data.

The invention discloses an optical remote sensing image segmentation method based on multi-scale lightweight hole convolution, and mainly solves the problems of large storage space occupied by a network and poor image segmentation effect in the prior art. The method comprises the steps of acquiring optical remote sensing image data, and dividing a training sample set and a test sample set; constructing a multi-scale lightweight hole convolution network formed by cascading a feature extraction lower sampling sub-network, a bottom layer sub-network and an image recovery upper sampling sub-network; training the constructed multi-scale lightweight hole convolutional network by using the training sample set; and inputting the test sample set into the trained multi-scale lightweight hole convolutional network for testing to obtain a segmentation result of the optical remote sensing image. According to the method, the storage space occupied by the segmentation network is reduced, the segmentation precision of the optical remote sensing image is improved, and the method can be used for land planning management, vegetation resource investigation and environment monitoring.

The invention provides a target fruit instance segmentation method and system, which belong to the technical field of computer vision. The method comprises steps of for an acquired orchard environment image, processing the orchard environment image by using a pre-trained segmentation model to obtain an identification segmentation result, wherein the pre-trained segmentation model is obtained by training a training set, and the training set comprises a plurality of orchard environment images and labels for labeling target fruits in the images, and when the orchard environment image is processed by using a pre-trained segmentation model, conducting semantic category recognition and mask segmentation on the extracted features to obtain a target fruit instance segmentation result. According to the method, the dependence of a model on an anchor frame is avoided, the complexity of the model is reduced, and mask annotation is independently used for instance segmentation tasks in an end-to-end mode to optimize a network; the method does not depend on a model detection frame, directly obtains the pixel segmentation result of the instance, improves the target fruit recognition and segmentation speed, is high in robustness and real-time performance, and reduces the missing detection rate and the false detection rate.

The invention discloses a streetscape image semantic segmentation system and segmentation method, electronic equipment and a computer readable medium. The segmentation method comprises the following steps: step 1, acquiring a streetscape image and carrying out preprocessing and data enhancement on the streetscape image; step 2, encoding the streetscape image into an output feature map by using an encoder; step 3, collecting features of the last three output feature maps by using a multi-level feature combined up-sampling module, and fusing the features to obtain a second output feature map; 4, converting the second output feature map into a third output feature map; 5, inputting the third output feature map into a convolution classifier to obtain a semantic segmentation feature value; step 6, performing end-to-end training by using a back propagation algorithm to obtain a streetscape image semantic segmentation model; and 7, performing semantic segmentation on the streetscape image by using the streetscape image semantic segmentation model. According to the method, under the condition that semantic segmentation precision is not reduced, the speed of network segmentation is increased, and the real-time response capability of the method in application is enhanced.

The invention discloses an interactive image segmentation method for reducing manual intervention. The interactive image segmentation method for reducing the manual intervention comprises the following steps of: firstly, extracting a small zone, which takes a boundary point for user input as a centre, and computing a mean gray value of the small zone; secondary, filtering an original image to be segmented with the mean gray value as a template so as to enhance the zone with the similar gray value and attenuate other zones; thirdly, extracting a border curve of the filtered image by a Canny operator so as to structure an energy consumption function capable of attenuating the boundary of a non-target object; fourthly, redefining an optimal path between two points as a path with minimum average energy consumption so as to make an Live-wire curve between the two points become a lasso with adjustable elasticity; and at last, computing a shortest path between two user input points as a boundary of a target object between the two points by using a Dijkstra algorithm.

The embodiment of the invention provides a hip joint CT image segmentation method and device, a storage medium and computer equipment. The method provided by the embodiment of the invention comprisesthe steps of segmenting a third hip joint CT image of a hip joint bone cortex area generated in advance through a graph cutting algorithm, and a left side pelvis area, a right side pelvis area, a leftfemur area and a right femur area are segmented; and adding the left-side pelvis area and the right-side pelvis area to obtain a pelvis area. According to the method of the invention, the image segmentation algorithm is improved in the hip joint CT image segmentation process, and the segmentation speed and segmentation accuracy of the hip joint CT image can be improved.

The invention discloses an attention mechanism-based lightweight semantic segmentation model construction method, which is applied to the technical field of image processing, and a training set is formed by giving an image I and a corresponding real label graph GT. The method comprises the steps of step 1, establishing a model; step 2, model training; and step 3, model testing: inputting a test set image into the trained network model to obtain a test result. According to the invention, the image segmentation accuracy and segmentation speed are improved; the segmentation process is not easy to over-fit; efficiency is high, and actual deployment is facilitated; and under the condition that the annotation data is insufficient, the annotation data is quickly trained, so that the performance is further improved.

The invention relates to a three-dimensional point cloud instance segmentation method, a three-dimensional point cloud instance segmentation system and electronic equipment. The method comprises the steps of a, inputting point cloud data into a point cloud instance segmentation model, performing feature extraction on the point cloud data by the segmentation model, and outputting semantic segmentation tags of the point cloud data and a high-dimensional vector of each point; b, predicting the object category of each point, and embedding the points into a high-dimensional vector; and c, after vector embedding is completed, predicting the'seed property 'of each point through a seed point selection network, and selecting a better seed point as a reference point to generate an instance to obtainan instance label of each point. According to the invention, a seed point selection network is added; according to the embodiment of the invention, 'seed 'judgment is carried out on each point in thepoint cloud data, and then a better seed point is selected to generate the proposal, so that better instance segmentation is realized, an obvious acceleration effect is achieved for post-processing of a network model, and the problems of low accuracy and low efficiency of a current point cloud instance segmentation technology are solved.

The invention discloses a noise image segmentation method based on an improved energy functional model, comprising the following steps: (1) inputting an original image; (2) inputting the original image in the step (1) denoised by using an energy functional model based on a non-convex functional, and the denoised smooth image is obtained; 3) initializing that smooth image in the step (2); (4) According to variational method and Euler-Lagrange equation, the level set function of evolution is obtained. (5) extracting a zero level set according to the level set function obtained in the step (4); (6), solving that minimum value of the energy functional, judging whet the evolution stops or not, if the evolution stops, the evolution curve is the best edge position of the target, give the segmentation result, otherwise, the algorithm goes to the step 4 to continue. Compared with the prior art, the invention improves the energy functional model, realizes a noise image segmentation method basedon the improved energy functional model, reduces the occurrence of ladder phenomenon, and improves the segmentation speed and the segmentation precision of the model while removing noise of the image.