Video saliency detection method based on deep network

Abstract

The invention discloses a video saliency detection method based on a deep network, and relates to the field of image data processing. The video saliency detection method comprises the following steps:firstly, using the ResNet50 deep network to obtain spatial features, then extracting time and edge information to jointly obtain a saliency prediction result graph, and completing video saliency detection based on the deep network, comprising the following steps: inputting a video frame I, and performing preprocessing; extracting an initial spatial feature map S of the video frame I'; obtaining aspatial feature map Sfinal of five scales; obtaining a feature map F; obtaining a rough space-time saliency map YST and an edge contour map Et of the saliency object; obtaining a final saliency prediction result graph Yfinal; calculating the loss of the input video frame I, and completing the video saliency detection based on the deep network. According to the method, the defects of incomplete salient target detection and inaccurate algorithm detection when foreground and background colors are similar in video saliency detection in the prior art are overcome.

Description

technical field [0001] The technical solution of the present invention relates to the field of image data processing, in particular to a video saliency detection method based on a deep network. Background technique [0002] Video saliency detection aims to extract the regions of most interest to human eyes in consecutive video frames. Specifically, it is one of the key technologies in the field of computer vision to use the computer to simulate the visual attention mechanism of the human eye and extract the area of ​​interest to the human eye from the video frame. [0003] Most of the traditional video saliency detection methods are based on low-level handcrafted features (such as color, texture, etc.), which are typical heuristic methods with slow speed (due to time-consuming optical flow calculation) and low prediction accuracy (due to limited characterization of low-level features). In recent years, deep neural networks have been applied to the field of video saliency d...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is: provide a video saliency detection method based on a deep network, the method is to use the ResNet50 deep network to get the spatial features, and then extract the time and edge information to jointly obtain the saliency prediction result map, and complete The video saliency detection based on the deep network overcomes the shortcomings of incomplete salient object detection and inaccurate algorithm detection when the foreground and background colors are similar in the prior art video saliency detection

Images