Dehazing images for better visibility 

Abstract

Image restoration is the key technique which is used to make certain that human beings as well as the critical and automated computer vision applications get quality image,free of any distortion.Haze is one of many distortions and impurifications faced by images captured through cameras.To accomplish the task of restoring degraded image many methods are presented.But these methods have shortcomings and ineffectiveness in itself.Here a SCM(Suppress and Complement Model) is proposed to dehaze variety of single hazy images .The proposed model uses the concept of gradually eliminating of unwanted element( haze).It also preserves and enhances the details of image like color priors,contrast and edges.During whole process ,focus is not only for dehazing but also uphold the originality of the image.Fast local laplacian filter (LLF) and rolling filter (RF) are used for diminishing haziness of the image as well as coordinated enhancing mechanism to preserve the details of the image.Due to its dynamic behaviour proposed method removes haze comparatively better than many previously available techniques. In this work,we have also proposed a method that is dynamic and agile in the sense that it can handle many variety of images.The contribution and quality of our work is to estimate and enhancement of unknown parameters of hazy model derived from atmospheric scattering model.This improvement is continuous and without defined boundary in the whole process.Due to its nature of work,this method is named as agile mechanism.Tree filter (TF) and joint weighted median (JWM) filter are materialized for smoothing and refinement at different stages of method.These refined and well tuned parameters are utilized to restore a quality haze free image. The quantitative metrics FADE(Fog Aware Density Evaluator), recovered edges (e),count of extreme pixel value change( ),and local contrast restoration (^r ) are used for comparison of proposed schemes with already presented techniques Qualitative comparison is done against visually faced issues in due course of dehazing like halo artefact,over saturation, darkness,white spikes around objects and better handling of sky area.Results achieved both quantitatively and qualitatively endorses that proposed scheme surmount previously available state-of-the-art dehazing algorithms.