Image collaborative saliency detection method based on multi-layer convolution feature fusion

Abstract

The invention discloses an image collaborative saliency detection method based on multi-layer convolution feature fusion. The image collaborative saliency detection method based on multi-layer convolution feature fusion includes the steps: 1) processing an image data set, including unifying the size, and selecting a collaborative image group for each image according to an sequence selection rule;2) constructing a deep learning network for collaborative saliency detection, inputting an image and its collaborative image group, performing extraction of multi-layer convolution features, extraction of collaborative features, fusion of multiple dimensional features, and fusion of multiple dimensional saliency images, and then obtaining a collaborative saliency image of the input image; 3) inputting the processed training data in the step 1) into the deep learning network constructed in the step 2) to train until network convergence can obtain a trained network model; and 4) performing an experiment on the test data set by using the trained network model in the step 3), and obtaining a plurality of collaborative saliency images by means of one input image and its multiple collaborative image groups, adding the input image and the multiple collaborative image groups the averages and averaging the input image and the multiple collaborative image groups to obtain a final collaborative saliency image of the input image.

Description

technical field [0001] The invention relates to an image co-saliency detection method, in particular to an image co-saliency detection method based on multi-layer convolution feature fusion, which aims to detect common salient objects from a set of images with common salient objects. Background technique [0002] With the development of media digitization and network technology, massive images and videos are produced every moment. People enjoy the rich visual information and fast and convenient interaction methods brought by images and videos. At the same time, the demand for personalized understanding and operation of these multimedia resources is becoming more and more urgent. Visual saliency detection models, which mimic the visual attention mechanism of the human eye to automatically capture striking objects in a scene, have attracted intense interest in the academic community. With the accumulation of similar images with common objects, finding common salient objects f...

AI Technical Summary

Problems solved by technology

However, this model only extracts the feature map of the last convolutional layer, and only considers the collaborative informati

Images