Self-Guided Image Dehazing Using Progressive Feature Fusion.

Area of Science:

• Computer Vision
• Image Processing

Background:

• Haze significantly degrades image quality, posing challenges for computer vision tasks.
• Existing image dehazing methods often struggle with preserving details and removing atmospheric light effectively.

Purpose of the Study:

• To develop an effective image dehazing algorithm leveraging self-guidance from the input hazy image.
• To improve haze removal by utilizing structural information from an intermediate pre-dehazed image.

Main Methods:

• A deep pre-dehazer generates a reference image with clear structures.
• A progressive feature fusion module integrates features from the hazy and reference images.
• An image restoration module restores the clear image using fused features, trained end-to-end.

Main Results:

• The proposed algorithm effectively removes haze while preserving image details.
• Experimental results demonstrate superior performance compared to state-of-the-art dehazing methods.
• The algorithm shows strong performance on benchmark datasets and real-world hazy images.

Conclusions:

• The proposed self-guided image dehazing algorithm offers an effective solution for haze removal.
• The integration of a deep pre-dehazer and progressive feature fusion significantly enhances dehazing performance.
• This method provides a robust approach for restoring clear images from hazy inputs.