Research On Video Dehazing And Its Quantitative Evaluation Mechanism

Owing to the attraction of the complex medium environment such as haze and dust, the quality of the obtained video is degraded badly. It makes image restoration work the degraded image to be a scientific problem urgent to be solved.This work focuses on video dehazing and its quantitative evaluation. Given that a hazy image in a natural scene generally exhibits low contrast and chromatic distortion, we proposed an image dehazing method using visual information loss prior. For the lack of an effective quantitative evaluation mechanism to the current image restoration techniques, we proposed a quality assessment framework for the performance ranking of image dehazing algorithms by employing prior features and learning methods. For the practical application, we developed the video dehazing platforms of software and hardware. Concentrating on the relevance theory of haze removal algorithms, this research is studied as follow:(1) Research on image haze removal using information loss prior. This paper proposes an image dehazing method that uses prior visual information loss. We ignore the transmission estimation and instead solve the optimization problems of the information loss function. First, the proposed method divides hazing images into three vision areas according to fog density. Second, the loss function, which is built based on the visible characteristics of hazy images, solves the local minimum transmission via the stochastic gradient descent method. Third, the divided dehazed areas are fused via multi-scale illuminance image segmentation with a linear filter. Fourth, the scene albedo is recovered by employing an atmospheric scattering model that uses global transmission. The proposed and existing dehazing methods are qualitatively and quantitatively evaluated to assess their image dehazing performance. The experimental results show that the proposed algorithm effectively removes haze from the degraded image and achieves higher-quality, halo-free, and detailed restorations than the existing dehazing methods. Compared with the state-of-the-art method, the proposed algorithm is more successful in recovering images from moderate to thick foggy areas and is faster in real-time dehazing applications.(2) Research on priori features guided quantitative assessment of image dehazing. We find that comparison or ranking mechanism of evaluation procedure is more close to a classification process rather than just predicting an absolute quality score. For blind image quality assessment of image dehazing performance, the concept of PIB is proposed to normalize the prior feature values. In this framework, prior information is used to extract the inherently features of hazy image firstly. And then, we compared these prior features with PIB to eliminate the cognitive bias of individuals. Finally, training a 3-grade classification network circularly with normalized features until the iteration ends and updating the PIB in every loop. The main contribution of our method is to formulate the evaluation of image dehaizng problem as a comparison framework by using classification methods, rather than previous prediction-based approaches for IQA.(3) Applied research on video haze removal algorithms. The software and hardware developments are based on the theoretical research on dehazing algorithms. In the testing software, we compared the state of the art haze removal methods and presented the files information of dehazing results in text display area. On the embedded hardware platform, the haze removal processing system is built based on TMS320DM642 processing chip, combining with CCD camera, DSP processing chip and video display. In the development process, we should cope with the contradiction between the computation and limited hardware resources. Finally, this system can be employed to deal with the task of real-time video haze removal in D1 resolution.