Research On Single Image Dehazing Algorithm

The technology of single image dehazing has been continuously developed.From eliminating the interference of haze visually through image enhancement,to researching the physical model and process of hazy image imaging,many achievements have been made in the field of image dehazing,among which image dehazing technology based on the physical model is widely used.However,in the actual image dehazing,there will still be poor effects such as incomplete dehazing,loss of edge details,color distortion and inapplicability to the sky area.Aiming at the problems existing in the dehazing algorithms,this thesis analyzes and studies the dehazing problems of hazy images based on the atmospheric scattering model.The main research work is as follows:1.Study the formation process of haze and the physical model of hazy image,and found that the accuracy of the estimation of atmospheric light value and transmittance value is the key to the ideal image dehazing effect,which provides a theoretical basis for the subsequent improvement of the algorithm.2.Research dehazing algorithms based on the haze-line prior and dark channel prior,and propose an improved dehazing algorithm,which combines these two prior principles to improve the methods of estimating the atmospheric light value and transmittance value separately.Aiming at the atmospheric light value estimation,based on the dark channel atmospheric light value estimation method,combined with the relationship between the HSV spatial haze concentration and the difference of brightness and saturation,the weight coefficient is set to adjust the atmospheric light value adaptively.The improved method fully considers the characteristic of atmospheric light value and has good anti-interference ability.To estimate the transmittance value,firstly use the dark channel prior algorithm to obtain the initial transmittance map;secondly find the transmittance value of the pixel point on each haze-line that is the farthest from the atmospheric light value,so as to obtain the transmittance of other pixels;finally the tolerance parameter and fast guided filtering are introduced to optimize the transmittance value.In the improved dehazing algorithm proposed in this thesis,the advantages of two dehazing algorithms based on prior information are fully combined,and finally a high-quality dehazed image is obtained.3.For images containing large areas of sky,the reason for the distortion of the sky area after dehazing is analyzed,and a sky area dehazing algorithm based on boundary-limited histogram equalization is proposed.Firstly the Canny operator in edge detection is combined with OTSU,and the sky area is accurately divided by the obtained gradient information and gray threshold;secondly according to the characteristics of the sky area,a boundary-limited histogram equalization algorithm is proposed to dehaze it;finally the two parts after the dehazing process are merged by image fusion technology to obtain a complete dehazed image.4.In order to verify the effectiveness of the dehazing algorithm proposed in this article,simulation experiments were carried out to compare with other classic dehazing algorithms,and the experimental results were evaluated from the perspectives of subjective vision and objective quantitative indicators.The experimental results show that the performance of the algorithm proposed in this article is superior.