Automatic single-image-based rain streaks removal via image decomposition.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Video rain removal is well-studied, but single-image rain removal remains challenging due to the lack of temporal information.
• Existing methods often struggle to preserve image details during the de-raining process.

Purpose of the Study:

• To develop an effective framework for single-image rain removal.
• To address the limitations of existing methods in handling single images by leveraging image decomposition techniques.

Main Methods:

• The proposed method formulates rain removal as an image decomposition problem using morphological component analysis.
• Images are first decomposed into low- and high-frequency (HF) components using a bilateral filter.
• The HF component is further decomposed into rain and non-rain components via dictionary learning and sparse coding.

Main Results:

• The proposed algorithm successfully removes the rain component from single images.
• The method effectively preserves original image details during the de-raining process.
• Experimental results validate the efficacy of the single-image de-raining approach.

Conclusions:

• The developed framework provides an effective solution for single-image rain removal.
• The image decomposition approach, combining bilateral filtering, dictionary learning, and sparse coding, is robust.
• This work advances the field of image de-raining by tackling the challenging single-image scenario.