The present invention discloses an image processing method and device. One concrete embodiment of the method comprises a step of extracting the characteristic of an image to be processed, and obtaining the first characteristic image of the image, a step of generating the attention image of the above image based on the first characteristic image, a step of fusing the above attention image and the above first characteristic image, and obtaining a fusion image, and a step of extracting the characteristic of the image again based on the fusion image. According to the embodiment, an attention mechanism is introduced into image processing, and the efficiency of obtaining information from an image is effectively improved.

Image processing method, device and electronic equipment

The invention is applicable to the technical field of biological identification, and provides a fingerprint identification method and a fingerprint identification device. The method comprises the following steps: minutiae of an input fingerprint pattern are extracted by a statistical method; and fingerprint matching is carried out according to the extracted minutiae to obtain a fingerprint identification result. According to the method provided by the embodiment of the invention, the directions of the minutiae are calculated based on statistical information, a statistically significant descriptor is added for the minutiae, and the steps of minutiae similarity calculation with the use of the descriptor and region matching with the use of orientation field and gradient field information of an overlap region are added in the matching process. Therefore, instability and poor specificity of fingerprint feature expression in a traditional algorithm are avoided, and the accuracy of fingerprint identification is improved.

Fingerprint identification method and fingerprint identification device

The invention provides a significant object detecting method based on a graph structure and boundary apriority. The method has the main thought that superpixels on the upper boundary, the lower boundary, the left boundary and the right boundary of an image are used as background apriority (boundary apriority), and a geodesic line distance from each superpixel in the image to the boundaries is obtained through building the graph structure and is utilized for measuring the significance. Firstly, input images are subjected to over-segmentation into superpixels; secondly, each superpixel is used as a vertex of the graph structure, in addition, four virtual vertices are added, three kinds of edges and weights of the edges are defined, and the complete graph structure is built; next, the shortest distance from each vertex to the four virtual vertices is calculated, and four significant graphs are obtained; finally, the final significant graph is output through one fusion strategy. The significant object detecting method has the advantages that the obtained significant graph can uniformly protrude significant objects in the image, particularly, the background can be perfectly inhabited, and the subsequent processing such as image retrieval and target tracking can be favorably carried out.

Significant object detecting method based on graph structure and boundary apriority

The invention discloses a mapping dictionary learning based cross-modality hashing search method. The mapping dictionary learning based cross-modality hashing search method includes: collecting image samples and text samples through a network, establishing a cross-media search data set, and dividing the data set into a training set and a test set; extracting features of the images and the texts through a BOW algorithm; learning a sharing sub space for an image mode and a text mode through mapping dictionary learning, and learning hashing functions of each mode at the same time; minimizing quantization errors by learning an orthogonal rotation matrix; calculating hashing codes of test samples through the hashing functions of the image mode and the text mode and the orthogonal rotation matrix; and using test samples of one mode as query samples, using the training set of the other mode as a searched data set, calculating a Hamming distance between the query samples and the searched samples and performing sequencing, and returning samples on top. The search method is high in search accurate rate, can be easily applied to large-scale data sets, is simple to implement, and has wide application prospect and huge market value.

Mapping dictionary learning based cross-modality hashing search method

Systems and methods for determining a breast region in a medical image are provided. The methods may include obtaining a first image relating to the breast region, determining a first region including a first plurality of pixels in the breast region, determining a second region including a second plurality of pixels relating to an edge of the breast region, and determining the breast region by combining the first region and the second region. The second plurality of pixels may include at least a portion of the first plurality of pixels.

System and method for determining a breast region in a medical image

Improved visual attention model-based method of natural scene object detection

The invention discloses an image perceptual hashing method based on two-sided random projection, which mainly solves the problem of quick searching of massive image data. The method comprises the following steps of (1) preprocessing original image data; (2) obtaining a two-sided random projection matrix; (3) carrying out low-rank approximating; (4) updating a low-rank matrix; (5) judging if the number of iteration times of the updated low-rank matrix is maximum or not; (6) grouping projection vectors; (7) orthogonalizing the projection vectors; (8) obtaining hashing codes; (9) obtaining a hamming distance; and (10) outputting test results. The hashing method has the advantages that the better projection vector can be obtained, the effective hashing codes can be obtained, the memory consumption is reduced, the searching time is saved, the comprehensive performance of precision ratio and recall ratio of image searching can be improved, and the hashing method is applied to the image searching services of electronic businesses and mobile terminal equipment.

Image perceptual hashing method based on two-sided random projection

The invention discloses an image retrieval method based on local locality preserving iterative quantization hash and mainly solves the problems that large-scale image retrieval has high memory usage rate and low retrieval performance. The method includes the steps of 1, extracting and normalizing original image features; 2, subjecting normalization data to principal components analysis to obtain a low-dimensional normalization data matrix; 3, dividing low-dimensional normalization data into training data and testing data; 4, construction a neighbor graph of image training data matrixes to obtain neighborhood relation of the image training data; 5, with the neighborhood relation as a constraint, updating a rotation matrix by iterative quantization; 6, acquiring a hash code of the image training data and that of the image testing data according to the rotation matrix; and 7, acquiring retrieval results according to a Hamming distance between the hash code of the training data and that of the testing data. The method has the advantages that memory consumption is lowered, image retrieval performance is improved, and the method is applicable to image retrieval services for mobile equipment, the internet of things and E-commerce.

Image retrieval method based on local locality preserving iterative quantization hash

The invention discloses a breast lump image feature extraction method based on edge neighborhood weighing The breast lump image feature extraction method mainly solves the problem that in the prior art, extracted features do not contain the edge neighborhood local features of a breast lump The breast lump image feature extraction method comprises the steps of (1) inputting an image, (2) adjusting the size of the breast lump image which is input, (3) extracting a lump edge, (4) determining the number of inner indentation pixel points and the number of outer extension pixel points, (5) determining the inner region of a lump which undergoes inner indentation, (6) determining the inner region of the lump which undergoes outer extension, (7) obtaining an edge neighborhood image of the breast lump, (8) obtaining weighing values, (9) extracting scale invariant features, (10) extracting word bag features and (11) obtaining features of the breast lump image which undergoes edge neighborhood weighing By means of the breast lump image feature extraction method, expression of the features of the breast lump image are more robust, the image features are expressed more effectively, the benign and malignant classification accuracy of lumps is improved, and therefore doctors in the radiology department can be assisted in conducting medical diagnosis

Breast lump image feature extraction method based on edge neighborhood weighing

The invention discloses a cross-model Hash retrieval method on the basis of cooperative training and matrix decomposition. By the aid of the cross-model Hash retrieval method, the similarity between models and the internal similarity of the models can be effectively constrained for unlabeled cross-model data. The cross-model Hash retrieval method includes implementation steps of acquiring original data and carrying out normalization processing on the original data; carrying out cooperative training to obtain constraints between the models; acquiring internal constraints of the models by the aid of neighbor relations; decomposing training data matrixes and adding the constraints between the models and the internal constraints of the models into the training data matrixes to obtain objective functions; carrying out alternate iteration to obtain expressions of basis matrixes, coefficient matrixes and projection matrixes; carrying out quantization to obtain Hash codes of training data sets and test data sets; computing the Hamming distances between every two Hash codes of the data sets; sorting the Hamming distances to obtain retrieval results. The cross-model Hash retrieval method has the advantages that constraints on the similarity between the models of the cross-model data can be obtained by the aid of cooperative training processes, accordingly, the image and text mutual retrieval performance can be improved, and the cross-model Hash retrieval method can be used for picture and text mutual search service of mobile equipment, internets of things and electronic commerce.

Deng Cheng

Papers

Improved visual attention model-based method of natural scene object detection

Image perceptual hashing method based on two-sided random projection

Image retrieval method based on local locality preserving iterative quantization hash

Breast lump image feature extraction method based on edge neighborhood weighing

Matrix decomposition cross-model Hash retrieval method on basis of cooperative training