Visual saliency algorithm based on motion characteristic of object in video

Abstract

The invention discloses a visual saliency algorithm based on the motion characteristic of an object in video. The visual saliency algorithm specifically comprises the following steps of: (1) calculating a motion vector of a brightness component of a current frame by using a block match-based motion estimation calculation method; (2) acquiring the motion vector from which the mean is removed; (3) performing Gaussian filter on the motion vector after the mean is removed; (4) calculating squares of components in the horizontal direction and components in the vertical direction of the motion vector respectively to obtain visual saliency maps of the horizontal direction and the vertical direction; and (5) acquiring a final visual saliency map. According to the method, when motion factors of the object in the video are considered, the motion saliency is extracted according to the local motion vector characteristic of the saliency object in the video by reducing the influence of the global motion of the object in the video on the visual saliency, so that the accuracy of the visual saliency algorithm can be effectively improved; and the algorithm provided by the invention has low complexity.

Description

technical field [0001] The invention relates to a visual saliency algorithm based on motion characteristics of objects in video, and belongs to the technical fields of computer vision and image processing. Background technique [0002] With the development of information technology, the amount of video information that people are exposed to in life is increasing. How to efficiently extract salient objects in videos has attracted more and more researchers' attention. Visual saliency has a wide range of applications in video signal processing. Applications, such as video retrieval, video compression, video surveillance, video tracking and other fields. [0003] In terms of video retrieval, since the amount of video data is very large, extracting salient objects in the video and using them as video features can effectively improve the accuracy of retrieval. [0004] In terms of video compression, since the video resolution is getting higher and higher, efficient video compress...

AI Technical Summary

Problems solved by technology

Compared with the global motion of the object in the video, the local motion of the object in the video is more significant, while the global motion of the object in the video is less significant. If the visual salience of the global motion of the object in the video is not reduced in advance will reduce the accuracy of the saliency algorithm

Images