FastDeRain: A Novel Video Rain Streak Removal Method Using Directional Gradient Priors

Area of Science:

• Computer Vision
• Image Processing

Background:

• Rain streaks significantly degrade outdoor video quality, posing challenges for various vision systems.
• Existing methods often struggle with efficiency and accuracy in rain streak removal.

Purpose of the Study:

• To propose a novel and efficient video rain streak removal approach named FastDeRain.
• To leverage the distinct characteristics of rain streaks and clean videos in the gradient domain.

Main Methods:

• Utilizing the sparsity of rain streaks and the anisotropic spatial smoothness of clean videos in the gradient domain.
• Employing unidirectional Total Variation (TV) regularizers and a time-directional difference operator for spatial and temporal consistency.
• Solving the minimization model using a split augmented Lagrangian shrinkage algorithm (SALSA).

Main Results:

• The FastDeRain method demonstrates effectiveness and efficiency on both synthetic and real-world datasets.
• Quantitative performance measures show superior results compared to state-of-the-art methods, particularly in terms of running time.
• The proposed approach successfully removes rain streaks while preserving video quality.

Conclusions:

• FastDeRain offers a significant advancement in video rain streak removal technology.
• The method's efficiency and accuracy make it a valuable tool for outdoor vision systems.
• The approach provides a robust solution for enhancing video clarity in adverse weather conditions.