Optimization method for foreground segmentation problem

Abstract

The invention discloses an optimization method for a foreground segmentation problem, and in particular discloses a optimization method for a foreground segmentation result based on a confidence idea, which belongs to the field of image processing, and aims to optimize a foreground segmentation result and avoid the activity phenomenon in the foreground segmentation result. The optimization method comprises the following steps of: carrying out foreground segmentation by a foreground segmentation method; and based on the confidence idea, transmitting one-dimensional information in four directions in parallel in an image to strengthen the association among pixels, thus optimizing the foreground segmentation result. The optimization method can eliminate the cavity phenomenon in the foreground segmentation result very well.

Description

technical field [0001] The invention relates to the field of video processing, in particular to an optimization method for foreground segmentation results. Background technique [0002] The segmentation of moving objects in video is widely used in various intelligent systems related to video surveillance. The quality of segmentation directly affects the accuracy of a series of motion analysis such as detection, tracking, recognition, and understanding of the target of interest. [0003] At present, some commonly used methods for foreground segmentation include frame difference method, background subtraction, optical flow method, and background modeling method. These methods have their own advantages and disadvantages. High, the effect is poor in scenes with cluttered backgrounds or large lighting changes; the optical flow method has a large amount of calculation, poor real-time performance, and is sensitive to background lighting, which will cause large errors; Each pixel b...

AI Technical Summary

Problems solved by technology

[0003] At present, some commonly used methods for foreground segmentation include frame difference method, background subtraction, optical flow method, and background modeling method. These methods have their own advantages and disadvantages. High, the effect is poor in scenes with cluttered backgrounds or large lighting changes; the optical flow method has a large amount of calculation, poor real-time performance, and is sensitive to background lighting, which will cause large errors; Each pixel builds a model and can update the pixel model. This method works well in complex and changing environments; the above methods are widely used in the field of video processing, but these foreground segmentation methods do not take pixels into account when performing segmentation calculations. Correlation between, so there will be holes in the target object when performing foreground segmentation

Images