Integrated prediction and time-space domain transmission-based video saliency detection method

Abstract

The invention discloses an integrated prediction and time-space domain transmission-based video saliency detection method. The method comprises the following steps of: (1) for the current frame of a video, constructing a local time domain window which takes the current frame as a center; (2) calculating a required optical flow field (motion vector field) by utilizing an optical flow algorithm, carrying out super-pixel segmentation and finally extracting region features; (3) obtaining an integrated saliency prediction model by utilizing two frames of information (comprising a corresponding saliency map) before the current frame, and carrying out saliency prediction on the current frame; (4) carrying out forward time domain transmission on the current frame by utilizing the two frames of information, obtaining a corresponding rough time-space saliency map by acting the two frames of information on two frames after the current frame, and carrying out backward time domain transmission on the current frame; and (5) carrying out space domain transmission to obtain a time-space saliency map corresponding to the current frame. Non-restraint video results show that the method is capable of uniformly lighten and highlight salient moving objects, and effectively restraining the background at the same time.

Description

technical field [0001] The invention relates to the technical field of image and video processing, in particular to a video saliency detection method based on integrated prediction and time-space domain propagation. Background technique [0002] With the popularization of wearable devices, smart phones and tablet computers with camera and video recording functions, the acquisition and storage of video information has become easier and easier. A large number of unconstrained videos also brings new challenges to research fields such as image and video processing. In recent years, studies have shown that the human visual system can quickly locate the most eye-catching objects from complex scenes, and how to use computer technology to simulate the human eye's visual mechanism and use it to extract human eye interest areas in images and videos has also become a current issue. Research hotspot. In the past few decades, researchers have proposed numerous saliency models and applie...

AI Technical Summary

Problems solved by technology

The motion saliency measure used in this algorithm is used as the initial saliency map, but it often fails for unconstrained videos such as complex scenes

In summary, the existing spatio-temporal saliency models cannot effectively highlight salient moving objects and suppress backgrounds, especially for unconstrained videos, which also hinders the widespread application of video saliency detection methods.

Images