85 results about "Significance values" patented technology

Apparatus, systems, and methods can operate to provide efficient data clustering, data classification, and data compression. A method comprises training set of training instances can be processed to select a subset of size-1 patterns, initialize a weight of each size-1 pattern, include the size-1 patterns in classes in a model associated with the training set, and then include a set of top-k size-2 patterns in a way that provides an effective balance between local, class, and global significance patterns. A method comprises processing a dataset to compute an overall significance value of each size-2 pattern in each instance in the dataset, sort the size-2 patterns, and select the top-k size-2 patterns to be represented in clusters, which can be refined into a clustered hierarchy. A method comprises creating an uncompressed bitmap, reordering the bitmap, and compressing the bitmap. Additional apparatus, systems, and methods are disclosed.

The invention discloses a video significance detecting method based on area segmentation, wherein the method mainly settles a problem of low detecting accuracy by an existing video significance detecting method. The video saliency detecting method comprises the steps of 1, performing linear iteration clustering on video frames, thereby obtaining a super-pixel block, and extracting the static characteristic of the super-pixel block; 2, by means of a variational optical flow method, obtaining the dynamic characteristic of the super-pixel block; 3, fusing the static characteristic and the dynamic characteristic for obtaining a characteristic matrix, and performing K-means clustering on the characteristic matrix; 4, performing linear regression model training on each cluster, thereby obtaining a regression model; and 5, reconstructing a mapping relation between a test set sample and a obtaining the significance value of a test set super-pixel block, and furthermore obtaining the significance graph of a testing sequence. Compared with a traditional video significance algorithm, the video significance detecting method has advantages of improving characteristic space and time representation capability, and reducing effect of illumination to detecting effect. The video significance detecting method can be used for early-period preprocessing of video target tracking and video segmenting.

A method reconstructs a depth image encoded as a base layer bitstream, and a set of enhancement layer bitstreams. The base layer bitstream is decoded to produce pixels of a reconstructed base layer image corresponding to the depth image. Each enhancement layer bitstream is decoded in a low to high order to produces a reconstructed residual image. During the decoding of the enhancement layer bitstream, a context model is maintained using an edge map, and each enhancement layer bitstream is entropy decoded using the context model to determine a significance value corresponding to pixels of the reconstructed residual image and a sign bit for each significant pixel, and a pixel value of the reconstructed residual image is reconstructed according to the significance value, sign bit and an uncertainty interval. Then, the reconstructed residual images are added to the reconstructed base layer image to produce the reconstructed depth image.

The invention discloses an asset dependence relation-based information security risk evaluation method and system. The method comprises the following steps of acquiring an asset range in a to-be-evaluated information system, and dividing the asset into blocks; identifying threats, vulnerability and asset significance of the to-be-evaluated information system, and acquiring significance value of the asset, threat value of the threat, and a vulnerability value of the vulnerability; calculating to build a security dependence relation matrix of assets by using a dependence structure matrix, determining a risk conduction coefficient of the assets by using a Delphi method and building a risk conduction relation; and calculating interior risk value and an exterior risk value of each block and overall risk value of the to-be-evaluated information system according to the risk conduction relation and asset significance, threat value and vulnerability value. The method can acquire a weak link ofthe system more accurately, and acquire more reliable information security risk evaluation results.

The invention discloses a video image significance detection method based on dynamic color association, which comprises the following steps of: obtaining a static significance chart of the video image; extracting the optical flow vector field of a scene; performing preliminary classification to the optical flow vector field and putting the maximum classification block away; converting the video image to HSV (hue, saturation, value) color space from RGB (red, green, blue) color space; generating a color histogram according to the frequency of the corresponding color in the H vector of the HSV color space appearing in the input image; aiming at each vector in the effective classification block of the optical flow vector field, projecting the norm into corresponding zones of the color histogram to obtain the movement scale variable of each color zone; obtaining the dynamic significance value of each color and projecting to the initial image to generate a dynamic significance chart; and summing the dynamic significance chart and the static significance chart, thereby obtaining the final significance chart. The method disclosed by the invention can effectively bring the dynamic characteristic into the significance consideration range, and can obtain the result on the basis of the existing dynamic video test set, which is more excellent than the result of the traditional method.

The invention is suitable for the image processing technology field and provides a stereoscopic image significance detection method, device and equipment, and a storage medium. The method comprises the following steps of when the request of stereoscopic image significance detection is received, firstly, using the superpixel image segmentation algorithm to segment the color image and the depth image of the stereoscopic image; then, carrying out characteristic extraction on all pixel areas acquired after segmentation through the significance prediction network of a multi-scale area level, acquiring a significance value corresponding to all the pixel areas, processing a pixel area corresponding to each significance value according to each significance value, and acquiring the color significance image of the color image and the depth significance image of the depth image; and finally, through a space fusion network, fusing the color significance image and the depth significance image and generating the significance image of the stereoscopic image. The error of extracting the significance characteristic of the stereoscopic image is reduced and the significance detection accuracy of thestereoscopic image is increased.

The invention discloses an image fusion method based on brightness self-adaption and significance detection. The method comprises the following steps: collecting an infrared image and a visible lightimage, and respectively carrying out image preprocessing and image registration; determining a brightness level by using histograms of the infrared image and the visible light image after graying processing, counting image pixel significance values and calculating a brightness weight; carrying out image decomposition by utilizing a rolling guided filtering method; defining pixel saliency values, and combining the plurality of pixel saliency values into a saliency map; carrying out fusion of a base layer image and fusion of a detail layer image; and superposing the base layer fusion image and the detail layer fusion image to obtain a final image fusion result. According to the method, the image brightness is graded, the to-be-fused image is decomposed based on rolling guided filtering, global targets and details of the image are processed separately, the brightness weight and the saliency map are used for fusing the base layer image, the least square method is used for fusing the detaillayer image, and the image fusion effect is improved.

The present invention discloses a method for detecting synergy significance of an RGBD images based on multi-core enhancement and significant fusion. The method comprises: (1) dividing a group of color images with a common significant object into several areas and calculating a significance graph of a single RGBD image; (2) carrying out sample selection, and selecting a category with the highest synergy significance value in the optimal clustering results, taking areas in the category as the positive sample, and taking the areas with the significance value of a single image less than the threshold as the negative sample; (3) using the image random sampling method to generate different training sets to learn a plurality of different models, so as to obtain the synergy significance graphs based on the multi-core enhancement; and (4) carrying out linear fusion on the synergy significance graph based on the multi-core enhancement obtained in the step (3) and the basic significance graphs obtained in the step (2) to obtain fusion synergy significance graphs, evaluating the quality of each fusion synergy significance graph, and taking the quality evaluation as the weight to carry out adaptive fusion so as to obtain final synergy significance graphs of the RGBD images.

The embodiment of the invention discloses a method and device for extracting the dominant hue of a picture, and is used for realizing accurate extraction of the main hue of the picture. The method may include the following steps: smoothing an input picture to obtain a smoothed picture; calculating the vision significance of the input picture to obtain a vision significance value corresponding to each pixel point in the input picture; according to the pixel value of each pixel point in the smoothed picture, determining the color region in color space to which each pixel point in the smoothed picture belongs, the color space including a plurality of color regions; and using the vision significance value corresponding to each pixel point as a weighted value to obtain statistical values corresponding to the plurality of color regions respectively, and determining the main hue of the input picture according to the statistical values corresponding to the plurality of color regions respectively, the statistical value corresponding to each color region including the number of all the pixel points in the corresponding color region and the corresponding vision significance values.

A computerized data mining method for determining a prediction model for a dependent data mining variable based on a multitude of independent data mining variables, comprising replacement of the independent data mining variable with potential values from a global range by a multitude of independent local data mining variables, each relating to potential values from a sub-range of the global range. The method further comprises an initialization step and a looping sequence that includes steps of determining for every independent local data mining variable not yet reflected in the current prediction model a multitude of partial regression functions, determining for each of the partial regression functions a significance value, selecting the most significant partial regression function and the corresponding not yet reflected local data mining variable, and adding the most significant partial regression function to the current prediction model.

The invention provides an image significance target detection method based on maximum neighborhood and superpixel segmentation, which is used for solving the technical problem of low image saliency target detection accuracy in the prior art. The method comprises the following implementation steps: 1, carrying out super-pixel segmentation on a to-be-detected image; 2, counting the occurrence frequency of each color in the to-be-detected image; 3, performing color substitution on the to-be-detected image; 4, preprocessing the image after color substitution; 5, calculating an initial saliency image of the to-be-detected image; 6, determining significance values of the K super-pixel blocks; and 7, obtaining and outputting a final saliency image. According to the invention, the accuracy of image saliency target detection is improved, the image saliency targets can be consistent and highlighted, and the method can be used for an image preprocessing process in the field of computer vision.

The embodiment of the invention discloses a method and a device for cutting a video. The method comprises the following steps: obtaining a to-be-cut video, equally dividing the to-be-cut video into a plurality of video clips according to the preset duration, and arranging the video clips according to a video play order; obtaining a first video clip, which comprises a film title and a tail leader and is appointed by a user; carrying out a transcoding treatment on the first video clip and generating a plurality of video frames; calculating a motion significance value of each video frame according to a preset motion significance algorithm; calculating the difference value between the motion significance value of each frame and the motion significance value of the first frame on the basis of the motion significance value; determining the frame with the maximum absolute value of the difference value as a key frame; and cutting out the film title and the tail leader from the first video clip according to the position of the key frame in the video clip. Through the application of the method and the device for cutting the video, the video can be accurately cut; and the user experience is improved.

The invention provides an image significance detecting method and device. A target image is obtained and segmented to obtain image blocks; the significance values of target image blocks, which are image blocks beyond visual points in the image blocks, are calculated on the basis of an ith visual point, and a first visual point represents the image block positioned in the central area among the image blocks; the attraction force of each target image block for the ith visual point is calculated by utilizing the significance value of the target image block; an (i+1)th visual point is determined according to the attraction forces of the target image blocks for the ith visual point; and when the distance between the (i+1)th visual point and the ith visual point is lower than a preset value, the (i+1)th visual point is determined to be a significant area, (i+1) is assigned to i, and the step of calculating the signification values of the target image blocks on the basis of the ith visual point is turned to. According to the invention, the significant area is searched on the basis of the attraction force of visual points for the image from the aspect of a human visual mechanism, and the significance detection effect is higher.

The invention discloses a cross-domain vision search method based on significance detection. The method comprises the following steps: firstly, utilizing each super-pixel zone boundary connecting value to endow all the zones with different significance values, thereby acquiring a main object zone; performing multi-scale treatment on a target image in a database; extracting the feature of the main object zone, thereby acquiring a target image feature template; performing feature extraction on the main object zone of an inquired image and training a linear classifier; performing iteration training on a large quantity of negative samples, thereby acquiring an optimized inquired image feature template; finally, while searching, returning a zone with the highest response score as a final searching result according to the matching degree of each target image feature template and the inquired image feature template. According to the invention, the significance detection for the main zone is performed, so that the influence of a background zone on the searching result is reduced, the searching precision and efficiency can be effectively increased in the cross-domain vision search and the robustness is excellent.

The invention discloses an edge-based image significant region detection method. The method comprises the following steps of: 1, carrying out edge detection on an original image to obtain an ultrametric contour map UCM; 2, aiming at the UCM, obtaining two super-pixel divisions in two different scales through different threshold methods; 3, on a fine-granularity super-pixel division layer, considering the color contrast, space prior and boundary prior at the same time to obtain an initial significance value; 4, on the fine-granularity super-pixel division layer, establishing a non-directional diagram by taking each super-pixel as a node and taking edge intensities as edges, and obtaining background prior through calculating geodesic distances of different nodes on the diagram; and 5, jointly considering the initial significance value, the background prior and the consistency of different scales to obtain a final significance value. According to the method disclosed by the invention, uniform high-brightness significant objects or regions can be obtained, and benefits are brought to the applications such as image scaling and image segmentation.

The invention provides a significance human body regional detecting method based on regional integration. The method includes steps of taking an ultra-pixel regional unit as a node, orderly executing the adjoining integration and the overall integration according to an integration rule psi; orderly structuring an adjoining integration picture FORMULA and an overall integration picture FORMULA; combining the color contrast, space distance weight and background probability weight among the ultra-pixel regional units, and respectively calculating the significance value of the ultra-pixel zone unit at an initial integration picture FORMULA, the adjoining integration picture FORMULA, and the overall integration picture FORMULA; through a confidence genetic mechanism of a cellular automaton machine, updating three initial significance pictures, accumulating and summing the pictures, and acquiring an integrated significance pictures S. In the invention, the method is good for presenting the gradation information in the input image on the basis of the initial integration picture, the adjoining integration picture and the overall integration picture; meanwhile, the method sets the confidence matrix according to the background probability value, and is good for the optimized performance of the cellular automaton; finally, the integrated significance picture S including the significant human body zone is acquired.

The invention discloses a significance detection method based on background priors. The method comprises the steps that a color image and a depth image are taken as input; superpixel segmentation anddepth map quality evaluation are carried out on the input images through a preprocessing operation; based on depth selective difference and background priors, the initial significance value of each superpixel region is calculated; and finally an initial significance map is optimized through the minimization of a cost function to acquire the final significance detection result. According to the invention, the problem that a traditional significance detection method based on color images cannot detect objects with similar visual characteristics with the background is solved; the problem of misdetection, which is caused by the fact that a bottom background region cannot be ignored when significance detection is carried out simply by relying on depth information; is solved; and the method provided by the invention is suitable for significance detection with color images and depth images, has the advantages of good overall effect and high accuracy, and can effectively detect significant objects.

The invention provides an image repositioning method, device and terminal, and belongs to the technical field of computers. The method comprises the following steps: building a significance model of an original image; calculating the significance value of each pixel dot in the original image according to the significance model; repositioning a targeted image according to the significance value of each pixel dot, the original resolution of the original image and the target resolution of the targeted image. The device comprises a model building module, a value calculating module and an image positioning module. With the adoption of the image repositioning method, device and terminal, the problem that analyzing the content of the original image to obtain an important image dot is complex and leads to large calculating load is solved, and the effect of lowering down the calculating load is reached.

The invention discloses a CU significance-based QP selection algorithm. The CU significance-based QP selection algorithm comprises steps of selecting a video frame image, dividing the video frame image into multiple CU blocks, and calculating a significance value of each CU block; calculating an average significance value of all CU blocks in S10; and according to the significance value of each CUblock and an average significance value of all CU blocks in the selected video frame image, dynamically adjusting the QP value of the CU block, and acquiring a sensing QP value of each CU block. In the method, the QP corresponding to the CU is selected according to the significance value of the CU, so that the CU with higher significance can encode with smaller QP, and a highly significant regionhas higher compression quality.

The invention provides an image quality detection method, an apparatus, a computer device, and a storage medium, wherein the method comprises: Using the preset neural network model, the subjective quality score of the target image to be detected is predicted, The initial subjective quality score corresponding to the target image is generated, and the saliency detection is performed on the target image to determine the saliency value corresponding to the target image, and the initial subjective quality score corresponding to the target image is corrected by using the saliency value to determinethe final subjective quality score of the target image. In this method, the preset neural network model is used to predict the subjective quality score of the image, and the initial subjective quality score is obtained, which avoids the difference of artificial evaluation, improves the objectivity and reliability of image quality evaluation, and corrects the initial subjective quality score by significance value, and improves the accuracy of the subjective quality score.