Background subtraction via time continuity and texture consistency constraints

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Background subtraction is crucial for moving object detection.
• Challenges include illumination changes, dynamic backgrounds, and noise, leading to artifacts like holes, noise, and shadows.
• Existing methods struggle with complex scenarios.

Purpose of the Study:

• To propose a novel background update model for robust moving object detection.
• To address issues of holes, noise, and shadows in detected moving object areas.
• To improve background modeling and object localization accuracy.

Main Methods:

• Utilizes a novel background update model based on matrix factorization.
• Leverages temporal continuity of video content for background modeling.
• Incorporates Neighborhood Weighted Local Binary Pattern (NWLBP) for texture consistency and shadow suppression.

Main Results:

• The proposed method effectively solves problems of holes, noise, and shadows.
• NWLBP significantly suppresses background and foreground shadows.
• Experiments confirm accurate background model establishment and robust moving object localization.

Conclusions:

• The novel background update model demonstrates superior performance in moving object detection.
• The method is effective in complex scenarios with illumination changes and dynamic backgrounds.
• It offers a robust solution for accurate background modeling and object localization.