Video denoising based on a spatiotemporal Kalman-bilateral mixture model

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Low-light video capture often results in noisy image sequences.
• Effective video denoising is crucial for subsequent analysis and visual quality.

Purpose of the Study:

• To develop an advanced video denoising method for low-light conditions.
• To leverage spatiotemporal correlations for improved noise reduction.

Main Methods:

• A spatiotemporal Kalman-bilateral mixture model is proposed.
• Motion estimation using block-matching is performed on video frames.
• Kalman filtering (temporal) and bilateral filtering (spatial) are applied, with results weighted for final denoising.

Main Results:

• The proposed method achieves competitive performance in video denoising.
• Evaluations using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM) confirm effectiveness.
• The method successfully reduces noise in low-light video sequences.

Conclusions:

• The spatiotemporal Kalman-bilateral mixture model offers a robust solution for low-light video denoising.
• Combining temporal and spatial filtering provides superior denoising results.
• The proposed algorithm demonstrates significant potential for practical applications.