Image Haze Removal Using Hybrid Advanced Image Restoration & Image Enhancement Techniques

Abstract

Low contrast and blurring issues can influence the perception and recognition of images taken in low and poor light situations. Haze in the atmosphere also causes the same consequences. Enhancing and restoring such low-contrast images is therefore crucial. Numerous image algorithms have already been used to remove haze from images taken in a low-light, foggy setting. However, these algorithms did not dehaze and enhanced well using the current haze removal techniques. Based on this rationale, a new hybrid mechanism by fusing two methodologies, Dark Channel Prior (DCP) and Single Scale Retinex (SSR) technique, is proposed in this research work. DCP responds well to image restoration and Haze removal, while SSR performs post-enhancement. The hybrid technique overcame the problems occurring with the earlier techniques, e.g., high noise in images, low contrast, large haze gradient, low entropy, Halo Artifacts, detail loss, image blurring, edge preservation, etc. The calculations of the following parameters have been used to assess the effectiveness of the hybrid strategy, e.g., Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Mean square error(MSE), Haze improvement index or Visibility Metric (VM). The generalized and image processing toolbox of MATLAB 2016a is used to implement the proposed methods.