Image dehazing using polarization effects of objects and airlight

Area of Science:

• Computer Vision
• Image Processing
• Optics

Background:

• Polarization analysis is valuable for image dehazing.
• Existing methods assume polarization originates solely from airlight, which is often inaccurate.
• Both airlight and object radiance contribute to polarization in hazy conditions.

Purpose of the Study:

• To develop a new polarization hazy imaging model considering joint polarization effects.
• To introduce an effective method for synthesizing optimal polarized-difference (PD) images.
• To propose a decorrelation-based scheme for estimating object polarization and recovering haze-free images.

Main Methods:

• A new polarization hazy imaging model incorporating airlight and object radiance.
• Synthesis of optimal polarized-difference (PD) images.
• A decorrelation-based scheme to estimate object polarization from polarized images.
• Haze-free image recovery using the proposed model.

Main Results:

• Experimental results confirm the effectiveness of the new dehazing scheme.
• The method successfully recovers haze-free images.
• The approach provides valuable object polarization properties as a byproduct.

Conclusions:

• The proposed polarization hazy imaging model and dehazing method are effective.
• The technique overcomes limitations of previous polarization-based approaches.
• The byproduct polarization information enables extended applications in computer vision.