Study On Image Dehazing Algorithms Based On Dark Channel Prior

Fog and haze formation are very common phenomena that are highly predominant in nature and our everyday lives.The conditions are even expected in sunny weather conditions where illumination would be expected to be abundant.This high prevalence is associated with the fact that the evaporation of water mist which leads to the formation of tiny droplets in the atmospheres still render the occurrence of mist highly probable.Low visibility in foggy days results in less contrasted and blurred images with color distortion which adversely affects and leads to the sub-optimal performances in image and video monitoring systems.However,due to the complexity of the problem itself,some of the existing algorithms and degradation models fail to provide optimal solutions to the problem of image restoration.The main goal of the study is the establishment of effective defogging schemes for images and the restoration of image clarity.We aimed to address the problem of haze image degradation towards an improvement of the performances of optical image processing and machine vision systems.Focusing on foggy image degradation,the causes of image degradation are explained in detail and the approaches of image enhancement and image restoration for defogging are introduced.In this thesis,some existing algorithms targeting image defogging are refined and improved,while some new ideas and methods are also put forward as follows:This paper proposes an effective method for single-frame image defogging calculation for various application scenarios.The single image dark channel defogging algorithm aims to solve the image blur problem in a fast and effective manner.The algorithm uses the dark channel priori theory to estimate the light intensity of the surrounding environment,thus achieving the purpose of removing noise effectively.This paper also proposes an image defogging method based on scene segmentation,in which the blurred image is divided into different parts according to the depth of the scene,and then the scene with different depth is defogged.The results show that this method can achieve high color fidelity of images.Moreover,the paper proposes a new single frame dehazing algorithm based on the single scale retinex(SSR)algorithm,combined with the theory of atmospheric scattering.Compared with the SSR algorithm,the proposed algorithm is shown to effectively take advantage of the abundance of information within an image and offers a clearer dehazing output within distant image scenes.Five methods were used to assess the defog effect,including information entropy,average gradient,edge strength,variance and color reducibility.Compared with SSR algorithm and histogram equalization algorithm,our algorithm shows much advantage in recovering the details and the colors from both the objective and subjective assessment.Moreover,an enhanced and advanced form of the improved Retinex theory-based dehazing algorithm is proposed.The first improvement is that it can achieve novel in the manner in which the dark channel prior is efficiently combined with the dark-channel prior into a single dehazing framework.Most important that proposed approach performs through implementation with an adaptive filter and allows for the dark channel features to be efficiently refined and boosted.Finally,experimental results proved that proposed algorithm is a strong candidate for real-time systems due to its capability to realize efficient dehazing at considerably rapid speeds.