Research on Video Detail Matching and Enhancement Algorithm Based on Bidirectional Feedback Mechanism

Huang Lin

Guangmian Technology (Nanchang) Co., Ltd.

Abstract：

The clarity and detail restoration capability of video content are of great significance for visual perception, video analysis, and augmented reality and other applications. Traditional video enhancement methods mostly rely on one-way feature extraction and enhancement strategies, which are difficult to fully restore local details while maintaining the global structure. This paper proposes a video detail matching and enhancement algorithm based on a bidirectional feedback mechanism. By combining forward prediction and reverse feedback, it achieves multi-scale information interaction and fine feature enhancement. In the algorithm design, spatio-temporal consistency constraints and residual learning mechanisms are introduced to improve the continuity and stability of details between video frames. Experimental results show that compared with existing mainstream enhancement algorithms, this method improves the PSNR and SSIM objective indicators by an average of 1.5-2 dB, and significantly enhances the texture details and edge clarity in terms of visual perception. This research provides an effective method for high-quality video enhancement, and has practical application value in video surveillance, post-production of films and television, and intelligent visual analysis.

Key Words：

video enhancement; bidirectional feedback; detail matching; residual learning; spatio-tmporal consistency