Effective dehazing of night-time images using open dark channel prior and wavelet transform

Existing night-time dehazing methods had been attempting to process light and non-light source regions based on dark channel prior (DCP). Since the bright and non-bright regions exhibit different features, the same daytime method cannot be applied to night images because light scatter from the multiple objects non-uniformly and DCP tends to over-estimate the depth of the scene making the image unrealistic. To overcome this limitation, wavelet decomposition was performed so that haze remains in the low occurrence region and noise in the high occurrence region and noise was removed by soft thresholding method. In the presented approach, the open DCP (ODCP) transmission map was computed for handling light source regions and estimated transmission was refined to enhance the texture in high-frequency part. Bilinear interpolation method of fast-guided filtering and recursive filter in the domain transform was used for edge preservation, enhancement of texture details and smoothness. The dehazed image was constructed by correlating the coefficients of low occurrence part recovered from haze and high occurrence component. The performance analysis was compared against state-of-the-art methods in terms of peak signal-to-noise ratio (PSNR) and Structural Similarity Index (SSIM).