Retinex-Based Laplacian Pyramid Method for Image Defogging

Abstract

To address the problem that the retinex algorithm cannot effectively enhance color and detail simultaneously, we propose a retinex-based laplacian pyramid method for image defogging. The method is implemented via MSRCR and laplacian pyramid, and it doesn't require additional hardware devices. The overall defogging process is composed of three vital parts: illumination color enhancement, detail of reflection component enhancement, and linear weighted fusion. Firstly, we add the gamma correction illumination back to reflection to achieve color enhancement. And then, the detail enhancement is achieved by the laplacian pyramid to process the reflection component. Finally, the detail enhanced image and color corrected image are used to reconstruct the clear image. Comparing with the state-of-the-art image defogging methods, the experimental results have shown that the contrast of defogged images can be effectively improved by the proposed method, and with better effects on both subjective and objective results. The contrast of images can be enhanced by the proposed method. Meanwhile, sufficient details can be preserved. Therefore, the proposed method is an effective image defogging method.