36 results about "Content-based image retrieval" patented technology

An image retrieval system performs both keyword-based and content-based image retrieval. A user interface allows a user to specify queries using a combination of keywords and examples images. Depending on the input query, the image retrieval system finds images with keywords that match the keywords in the query and / or images with similar low-level features, such as color, texture, and shape. The system ranks the images and returns them to the user. The user interface allows the user to identify images that are more relevant to the query, as well as images that are less or not relevant to the query. The user may alternatively elect to refine the search by selecting one example image from the result set and submitting its low-level features in a new query. The image retrieval system monitors the user feedback and uses it to refine any search efforts and to train itself for future search queries. In the described implementation, the image retrieval system seamlessly integrates feature-based relevance feedback and semantic-based relevance feedback.

A method is for automatically indexing multimedia data archives and categorizing the files held therein and also to a client / server architecture in an image retrieval system for content-based searching for relevant files in a particular format and having a particular file structure. The parsed files stored in a data archive managed by the document management system are subjected to a feature extraction algorithm. The features obtained are then used for producing a binary-coded inverted index which includes elements of at least two attributes and holds context information which is held in these files and is needed for content-based image retrieval. If new files or files with an extended or modified content are stored in the data archive, then it involves a parsing algorithm and also an algorithm for automatically extracting features of these files being executed under event control during every storage process, in order to extend the inverted index by individual attributes or in order to update particular elements of already existing index attributes.

The invention belongs to the technical fields of mode indentifying and image processing, and relates to an image matching method based on characteristic points. The image matching method based on characteristic points defines an annular corner code model according to the K neighboring structure of the characteristic points and applies the model to point mode matching. Similar annular corner codes not only describe the similarity of the space structures of two characteristic points, but also can be used for local mapping estimation. The similarity of the two characteristic points is confirmed by the similar length of the largest similar annular corner codes associated with the two characteristic points. The image matching method based on characteristic points carries out structure matching according to the similarity of the local space structures of the characteristic points, and uses the local mapping and clustering to carry out optimization matching. The optimization matching result is used for optimizing the mapping estimation. The method can be used for the image matching based on the characteristic points, image jointing and embedding, motion tracing, information fusion of medical single-mode and multi-mode images, as well as image searching based on contents.

The invention relates to an image searching feedback method based on contents, which comprises selecting a training sample from an image data base, training the training sample by utilizing a support vector machine to obtain a feature classification model, and classifying an image into a visual category; determining the visual category of the image submitted by a user according to the feature classification model, searching an image similar to the image submitted by the user, and returning a searching result; selecting feedback images in the searching result, marking images to be positive andnegative feedback images respectively according to the relevance between the feedback images and the image submitted by the user, and feeding back the mark result to a searching system; judging the accurate category of the image submitted by the user according to text keywords of the feedback images, category labels and a mapping table, searching the image similar to the image submitted by the user in the accurate category, and returning a secondary searching result. The image searching feedback method based on contents can fast and accurately locate the category the image submitted by the user belongs to and improve secondary searching accuracy.

The invention discloses a cross-media information retrieval system based on a mobile visual device, comprising: an image input module for inputting an image into the mobile visual device, a text input module for inputting text information related to the image , a text-based Internet image retrieval module for retrieving an image collection associated with the input text information and establishing a text-related image collection, for searching an image matching the input image in the text-related image collection, and Establish a content-based image retrieval module that generates visually relevant image sets and extract keywords from the web pages where the images in the visually relevant image sets are finally found, and then extract relevant online content and display the retrieval results. The first retrieval processing module . The invention combines text and image retrieval to realize accurate and fast cross-media and visual retrieval. The invention also discloses a cross-media information retrieval method based on the mobile visual device.

The invention provides an Internet-based design method of an artistic picture splicing system. The method refers to the content-based image retrieving technology in the image retrieving field and the image-synthesis-related method in the computer graphics, and according to input images, a plurality of images searched from the Internet are spliced into an artistic spliced image. Different from theconventional content-based image retrieving method, the method has the advantages that a brand-new energy function for retrieving based on shape and color is provided; a relatively ideal retrieving result is obtained; when the hierarchical relationship between subblocks is determined, three important clews used for judging the hierarchical relationship between the neighboring subblocks are provided and the topological sorting is introduced so as to generalize all local hierarchical relationships effectively to obtain an ideal general hierarchical relationship; and according to the input images, an artistic spliced image is obtained and the method is a brand-new attempt in the fields of recreation and art.

The invention discloses an image retrieval method based on group sparse feature selection. The image retrieval method includes a feature selection step and an image retrieval step. The feature selection step includes selecting image pairs, extracting multiple features, forming a characteristic difference matrix, establishing a group sparse logic regression model, utilizing an optimization algorithm to solve weight and selecting the optimum features. The image retrieval step includes extracting optimum features of all images in an image bank, forming an image feature bank, extracting and searching for image optimum features, carrying out similarity comparison, solving the largest similarity image sequence number and outputting retrieval images. According to the image retrieval method based on group sparse feature selection, a self-adaptation spectrum gradient algorithm is adopted to effectively solve the group sparse logic regression model, the rate of convergence is higher and the operating time is shorter. For a traditional image retrieval method based on contents, selection of the features is specified. For the image retrieval method based on group sparse feature selection, image retrieval and comparison are carried out by utilizing all training features, so under the same experiment condition, the image retrieval method based on group sparse feature selection has the advantages that the precision ratio of the images is improved obviously, and the efficiency of image retrieval is improved.

The invention relates to an image retrieval method based on deep convolution characteristic and semantic similarity measurement, and belongs to the correlation fields such as computer vision, image processing and image understanding. Firstly, aiming at an image set, convolution layer characteristics of each image are extracted through a trained deep convolution neutral network mode, the extractedconvolution layer characteristics are subjected to polymerization representation, the convolution layer characteristics are subjected to semantic description through an AFS framework, an image similarity measuring method based on semantic similarity is defined on the basis, the similarity of images in an image library are calculated, and finally the similarity is ranked to complete an image retrieval task. The problems that a traditional retrieval method based on bottom layer visual characteristics currently is in lack of semantic and low in accuracy can be effectively solved, and the actual requirements of users on image retrieval based on content are better met.

The invention discloses a contour detection stability judging method. The contour detection stability judging method is characterized by judging the contour detection stability according to the contour integrity of detected continuous multi-frame target objects and the position stability of comparison top points selected from contours. The contour detection stability judging method has the advantage of solving the problem of instable contour detection caused by hand vibration during a mobile terminal using process by a user. The invention also discloses an image searching method. The image searching method is characterized by firstly carrying out the contour detection of the target objects and judging the contour detection stability before carrying out image search, capturing a target gray image in the contours of the target objects which are judged to be stable, and uploading the target gray image to an image searching server side for carrying out the image search based on content. The image searching method has the advantages that the influence on an image searching result due to the image background is reduced, and the searching accuracy degree is increased. The contour detection stability judging method and the image searching method, disclosed by the invention, are particularly suitable for the filed of mobile cloud image search.

The invention discloses a content-based image retrieval method based on significance segmentation. The method comprises a training process and a testing process. The training process comprises the following steps that firstly, a visual vocabulary dictionary is established, specifically, all images in a training image set are subjected to analysis and processing sequentially, and the visual vocabulary dictionary is established for subsequent retrieval; secondly, target foreground images and background regional images are obtained through segmentation by means of visual significance features of the images; thirdly, color features and textural features are extracted from the target foreground images and the background regional images; and fourthly, statistics of visual vocabulary distribution histograms of all the images in the database are made on the basis of operation in the first step. The testing process comprises the fifth step that on the basis of operation in the four steps, tested images are retrieved. By adoption of the content-based image retrieval method based on significance segmentation, the number of semantic gaps can be effectively decreased, and the accuracy is high.

The invention discloses a content-based image retrieval method for unsupervised adversarial training. The method comprises the following steps of: 1, carrying out network construction, wherein an unsupervised adversarial network framework consists of a generation model and a discrimination model, and both the generation model and the discrimination model are formed by three-layer full-connection networks; 2, carrying out data preprocessing; 3, carrying out network training: 3.1, initializing generation model and discrimination model parameters by using random weight values, 3.2, training the generation model, and 3.3, training the discrimination model; and 4, carrying out precision test. The content-based image retrieval method for unsupervised adversarial training is relatively good in robustness and relatively low in training data requirement, and does not need a large amount of label information.

The invention relates to a system (100) for retrieving an image from image storage means, the system comprising: a retrieval unit (110) for retrieving a plurality of images from image storage means, on the basis of similarity between said images from image storage means and a query image, wherein the similarity is defined by a similarity function, and comprising a relevance unit (120) for computing values of relevance of images of the retrieved plurality of images on the basis of a value of a first attribute relating to the query image and values of a second attribute relating to the images of the retrieved plurality of images, and an update unit (130) for updating the similarity function on the basis of the computed values of relevance, and wherein the retrieval unit (110) is further adapted for retrieving the image from image storage means on the basis of the updated similarity function. The values of the first and second attributes are used by the retrieval unit to compute relevance values (e.g. ranks) of the plurality of images retrieved from image storage means. The user is thus relieved of the task of comparing the retrieved images and evaluating their relevance according to a user-defined criterion. Advantageously, the relevance values computed by the system are less dependent on, or independent of, user subjectivity. Optionally, the system may further comprise a prediction unit (140) for predicting the value of an attribute relating to the query image, on the basis of the image retrieved by the retrieval unit (110) using the similarity function updated by the update unit (130)

The invention provides a method for acquiring images of electric power equipment on the basis of multispectral sensor groups. The method includes steps of 1), acquiring image information, to be more specific, acquiring spectral information and site information by the aid of collection units comprising multisource sensors and carrying out data preprocessing on acquired information of the sensors bymeans of data conversion and fusion; 2), storing and managing the information, to be more specific, judging types of equipment images acquired from different sources and constructing image data storage structures; 3), analyzing information types, and to be more specific, providing extensible interfaces for general image operation such as image retrieval and image rough classification on basis oftexts and contents for follow-up image processing operation by the aid of image analysis interfaces; 4), outputting results, and to be more specific, judging and outputting the detection results by the aid of a data processor. The method has the advantages that image data can be efficiently utilized when multispectral images of power transformation equipment are analyzed, accordingly, safe and reliable operation of power grid equipment can be guaranteed, and the method has high theoretical and practical significance.

The invention belongs to the technical field of image detection, and relates to a composite image detecting method based on image search. The method comprises the following steps: inputting an image which is to be detected whether to be a composite image in a content-based image search system, searching, and selecting an image with the highest matching degree from the searched matched images; and adjusting the image to be detected and the searched image with the highest matching degree to the same size, subtracting, and performing the following judgments: a, if the subtracted image has a large area of black part or a small area of continuous black part and a large area of continuous white region, proving that at least one of the two images is processed; and b, if the subtracted image has a large area of white region and smaller and scattered black regions, judging that the relation between two images is small, i.e., the image is not processed. The detecting method is simple, intuitive, low in time complexity and easy for application.

The invention discloses a saliency detecting element evaluation method based on image retrieval. The method comprises the following steps: respectively taking a saliency map labeled by a user and a saliency map generated by T saliency detection algorithms to serve as a weighted color histogram of a weight calculation input image, and acquiring a retrieval image by utilizing an image retrieval method based on the content so as to respectively obtain the retrieval sequences L and Lk; respectively calculating similarity values of the retrieval sequences L and Lk, and sorting to obtain an evaluated sorting sequence X of the image retrieval application on the T saliency detection algorithms; respectively calculating evaluation values of the saliency detection evaluation method on the T saliency detection algorithms, and sorting to obtain an evaluated sorting sequence Xb of the saliency detection evaluation method on the T saliency detection algorithms; calculating the correlation between the sequences X and Xb; and taking the average value of the correlations of all input images in an image set as the evaluated value of the saliency detection evaluation method. The method disclosed by the invention contributes to selecting a proper and effective saliency detection evaluation method for the actual application.

The invention discloses an image retrieval method based on group sparse feature selection. The image retrieval method includes a feature selection step and an image retrieval step. The feature selection step includes selecting image pairs, extracting multiple features, forming a characteristic difference matrix, establishing a group sparse logic regression model, utilizing an optimization algorithm to solve weight and selecting the optimum features. The image retrieval step includes extracting optimum features of all images in an image bank, forming an image feature bank, extracting and searching for image optimum features, carrying out similarity comparison, solving the largest similarity image sequence number and outputting retrieval images. According to the image retrieval method based on group sparse feature selection, a self-adaptation spectrum gradient algorithm is adopted to effectively solve the group sparse logic regression model, the rate of convergence is higher and the operating time is shorter. For a traditional image retrieval method based on contents, selection of the features is specified. For the image retrieval method based on group sparse feature selection, image retrieval and comparison are carried out by utilizing all training features, so under the same experiment condition, the image retrieval method based on group sparse feature selection has the advantages that the precision ratio of the images is improved obviously, and the efficiency of image retrieval is improved.

The invention discloses an image retrieval method based on specific attributes of an HSV color space. Image retrieval is carried out by focusing on the advantages of unique attributes of an HSV colorspace and a visual perception mechanism, a novel image feature expression method of a cylinder feature histogram is provided, and color information and edge features in the HSV color space can be usedfor expressing image content. The cylinder characteristic histogram can be regarded as an important improvement of a histogram description method; the histogram can be used for expressing color, texture, shape and spatial characteristics; compared with the prior art, the method provided by the invention not only has the advantages of the histogram-based method, but also absorbs the advantages ofthe histogram-based method, and is more consistent with visual perception factors such as spatial layout, different color affiliation relationships and edge clues, so that the method is specially usedfor feature analysis and content-based image retrieval and has richer visual information, and the image retrieval performance can be obviously improved.

The invention discloses a clothing retrieval technology based on deep metric learning, and relates to the field of content-based image retrieval. The method comprises the following steps: 1) model design: performing feature embedding on a batch of clothing pictures through a convolutional neural network and a full connection layer; 2) a sampling method: selecting a training sample picture sample pair according to cosine similarity, namely known similarity information of positive and negative sample pairs; and 3) loss function: substituting the picture sample pair selected by the sampling method in the step 2) into the designed loss function to calculate the loss, and performing back propagation to train the model. Image features are extracted by using the design of a convolutional neural network rear full connection layer, the performance is good and simple, and the parameters are few; according to the sampling method, information carried by simple and difficult sample pairs can be effectively mined, and meanwhile, unlimited punishment is prevented from damaging the overall structure of an embedded space. The loss function can adjust the loss of the difficult sample pair along withthe training process so as to fully learn the information of the simple sample pair and the difficult sample pair.

The invention discloses an image importance detection method based on a photographic element, which comprises the following steps: (1) color analysis; (2) depth-of-field analysis; (3) structure analysis; and finally, outputting an importance diagram for reflecting the attention degree of each pixel of an image. The image importance detection method has the benefits that firstly, a method for reflecting the understanding to a scene when a photographer shoots the scene is provided, and through the method, the defect that a shallow depth-of-field image can not be processed through the existing importance diagram detection method can be effectively overcome; secondly, compared with the existing method, an provided image depth-of-field classification method has the advantages of high classification accuracy and quick classification speed; thirdly, the image importance detection method has wide application prospect and can be applied to the image retrieval based on contents, the automatic segmentation of image objects and the field of automatic monitoring and tracking of video objects and the like.

The invention discloses an image retrieval and classification method and system, a terminal and a storage medium. The method comprises the steps of obtaining a to-be-queried image; inputting the to-be-queried image into a target Hash classification network to obtain a predicted Hash code and a predicted image category corresponding to the to-be-queried image; and obtaining a target database Hash code, and determining a target image set corresponding to the to-be-queried image according to the target database Hash code and the prediction Hash code. The image category of the to-be-queried image and a group of similar images corresponding to the to-be-queried image are determined in a Hash code mode, the problem that the scene category of an image cannot be obtained by an existing image retrieval method based on content is solved, and further analysis and processing of the image are facilitated.

The invention relates to the technical field of image retrieval, and discloses an image retrieval method based on image information extraction and EMD distance improvement. The method comprises the following steps: a, carrying out dimension reduction processing on a to-be-retrieved image and an image library image; b, setting a network structure and extracting image semantic information after dimension reduction; c, improving the EMD distance; d, applying the improved EMD distance to image similarity measurement. According to the image retrieval method based on image information extraction and EMD distance improvement, the technical means are interdependent and indispensable, and as a whole, the technical purpose that the content-based image retrieval accuracy cannot be effectively improved when any one means exists is jointly achieved; in addition, according to the method, the image retrieval accuracy can be improved, and the applicability of a retrieval system to complex scenes can be enhanced.

The invention discloses a saliency detecting element evaluation method based on image retrieval. The method comprises the following steps: respectively taking a saliency map labeled by a user and a saliency map generated by T saliency detection algorithms to serve as a weighted color histogram of a weight calculation input image, and acquiring a retrieval image by utilizing an image retrieval method based on the content so as to respectively obtain the retrieval sequences L and Lk; respectively calculating similarity values of the retrieval sequences L and Lk, and sorting to obtain an evaluated sorting sequence X of the image retrieval application on the T saliency detection algorithms; respectively calculating evaluation values of the saliency detection evaluation method on the T saliency detection algorithms, and sorting to obtain an evaluated sorting sequence Xb of the saliency detection evaluation method on the T saliency detection algorithms; calculating the correlation between the sequences X and Xb; and taking the average value of the correlations of all input images in an image set as the evaluated value of the saliency detection evaluation method. The method disclosed by the invention contributes to selecting a proper and effective saliency detection evaluation method for the actual application.

The invention relates to an image retrieval method based on deep convolution features and semantic similarity measurement, and belongs to related fields such as computer vision, image processing, and image understanding. First, for the image set, use the trained deep convolutional neural network model to extract the convolutional layer features of each image, aggregate and represent the proposed convolutional layer features, and then use the AFS framework to describe them semantically. On this basis, an image similarity measurement method based on semantic similarity is defined, and the image similarity in the image database is calculated accordingly, and finally the image retrieval task is completed by sorting the similarity. The invention can effectively solve the problem of lack of semantics and low accuracy in the current traditional retrieval method based on the underlying visual features, and better meet the actual needs of users for content-based image retrieval.

A content-based image retrieval method for unsupervised confrontation training, the method includes the following steps: Step 1, network construction, the unsupervised confrontation network framework is composed of a generation model and a discrimination model. Both the generative model and the discriminant model are composed of a three-layer fully connected network; Step 2, data set preprocessing; Step 3, network training, the process is as follows: Step 3.1: Initialize the parameters of the generative model and discriminant model with random weights; Step 3.2 : training generative model; step 3.3: training discriminant model; step 4, accuracy test. The invention provides a content-based image retrieval method for unsupervised confrontation training with good robustness, low requirements for training data and no need for a large amount of labeling information.