Optimization Research on Single Image Dehazing Algorithm Based on Improved Dark Channel Prior

Abstract

Single image dehazing remains a challenging problem in computer vision due to the ill-posed nature of atmospheric scattering equations. Traditional dark channel prior methods demonstrate effectiveness in many scenarios but suffer from significant limitations in sky regions and bright objects. This research presents an optimized approach that addresses these deficiencies through enhanced transmission map estimation and refined atmospheric light calculation. The proposed algorithm integrates adaptive filtering mechanisms with improved boundary constraints to achieve superior dehazing performance. Experimental validation on synthetic and real-world datasets demonstrates substantial improvements in both quantitative metrics and visual quality compared to existing state-of-the-art methods. The optimized algorithm achieves an average PSNR improvement of 3.2 dB and SSIM enhancement of 0.15 while maintaining computational efficiency suitable for real-time applications. The research contributes novel enhancement strategies that advance the practical applicability of dark channel prior-based dehazing algorithms in diverse atmospheric conditions.