Learning-based visual attention prediction system and learning-based visual attention prediction method

Abstract

The invention relates to a learning-based visual attention prediction system and a learning-based visual attention prediction method. The method includes the steps: firstly, learning correlation relationship between fixation density and feature information via training; secondly, receiving a test video sequence with a plurality of test frames; thirdly, generating at least one tested feature map from each test frame on the basis of the feature information; and finally, according to the correlation relationship, corresponding each tested feature map to a saliency map used for expressing fixation strength of the corresponding test frame.

Description

technical field [0001] The present invention relates to a visual attention prediction system and method thereof, in particular to a visual attention prediction system and method based on learning-based video signals. Background technique [0002] Visual attention is an important feature of the human visual system, which helps our brains filter out excess visual information and allow our eyes to focus on specific areas of interest. Visual attention has been the subject of neuroscience, physiology, psychology, and human vision research. In addition to allowing us to understand the psychological aspects of visual attention, these studies can also be applied to the processing of video signals. [0003] Usually the gaze point in the film attracts the most attention. If the attention area in the film can be predicted, finer image processing or better coding procedure can be performed only on the film signal in this area. Traditional visual attention models consist of two parts: f...

AI Technical Summary

Problems solved by technology

However, due to improper weight assignment in the feature fusion procedure, or only capturing low-level feature information, such as color, orientation, etc., there will be visual inconsistencies between the predicted saliency map and the actual human gaze location. pairing problem

[0004] Since the traditional visual attention model cannot effectively predict the visual attention area, it is urgent to propose a novel visual attention prediction system and method to faithfully and easily predict the visual attention area

[0005] This shows that the above-mentioned existing relevant visual attention prediction system and method thereof obviously still have inconvenience and defects in methods and use, and need to be further improved urgently

In order to solve the above-mentioned problems, the relevant manufacturers have tried their best to find a solution, but no suitable design has been developed for a long time, and there is no suitable method to solve the above-mentioned problems in general methods and methods. This is obvious. It is a problem that relevant industry players are eager to solve

Images