IDRLP: Image Dehazing Using Region Line Prior.

Area of Science:

• Computer Vision
• Image Processing

Background:

• Hazy images degrade visual quality and hinder downstream tasks.
• Existing single image dehazing methods often struggle with robustness and efficiency.

Purpose of the Study:

• To propose a novel and ultra-robust single image dehazing method named IDRLP.
• To leverage a region line prior (RLP) for improved image restoration.

Main Methods:

• Region segmentation based on scene depth to observe a quasi-linear relationship (RLP).
• Combining RLP with the atmospheric scattering model (ASM) to derive a recovery formula (RF).
• Employing a 2D joint optimization function and a guided filter for accurate parameter estimation and noise reduction.

Main Results:

• The proposed IDRLP method demonstrates superior recovery quality compared to state-of-the-art techniques.
• IDRLP achieves ultra-high robustness and efficiency in single image dehazing.
• Extensive analysis validates the effectiveness of the RLP and joint optimization strategy.

Conclusions:

• IDRLP offers a significant advancement in single image dehazing.
• The RLP and joint optimization approach provides a robust and efficient solution.
• The method's superiority is confirmed through comprehensive evaluations.