Analytical system for attention of video signal

Abstract

The invention relates to the technical field of computer vision, in particular to an analytical system for the attention of a video signal, which comprises an attention feature extraction module based on multiple reference frames and an attention graph calculation module based on spatial continuity, wherein the attention feature extraction module based on multiple reference frames further comprises a luminance feature extraction submodule, a color feature extraction submodule, a direction feature extraction submodule, a motion feature extraction submodule based on multiple reference frames and a flicker feature extraction submodule; the attention graph calculation module based on the spatial continuity further comprises a spatial attention subgraph calculation submodule, a motion attention subgraph calculation submodule, a motion attention subgraph reinforcing submodule based on the spatial continuity, a flicker attention subgraph calculation submodule and an attention subgraph merging submodule. In the analytical system, the accuracy rate of the extraction in attention areas is improved substantially under the condition of not increasing calculation complexity obviously.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a system for analyzing the degree of attention of video signals. Background technique [0002] The visual attention model finds the area of ​​interest to the human eye in the image / video by simulating the human eye's perception mechanism of the scene image. At present, it has been widely used in human eye attention search, image segmentation, image compression, and adaptive image cropping. Cutting, video object tracking, video retrieval, video coding, 3D scene perspective and other fields. [0003] In 1980, Anne Treisman, a professor of psychology at Princeton University in the United States, concluded that color, direction and brightness are the most concerned features of the human visual system by simulating experiments on the human visual system. , which is an important theoretical basis for the human visual model. In 1998, Itti et al. conducted a pioneering study on...

AI Technical Summary

Problems solved by technology

However, the Itti model only uses two frames before and after to extract motion parameters, and the speed range of the objects that can be captured is limited. When the speed of the moving object is slow, it is easy to cause missed judgments.

In addition, its motion feature detection based on the frame difference method is easy to identify the interior of the moving object as a low attention area, resulting in misjudgment

[0004] In 2005, Wen-Huang Cheng in Document 4 aimed at the shortcomings of the Itti model’s poor handling of slow motion, segmented the video sequence, and used median filtering to obtain a time-segmented attention map, but this method is not suitable for severe motion The sequence of

In addition, in 2009, Chang Liu et al. used the statistical model of pixels to fuse the spatial attention degree and the temporal domain attention degree in 2009, so as to solve the problem of missed judgment when the moving object is slow, but the prerequisite is that the statistical The window must match the size of the foreground object, so it is not universal in practical applications

Images