Study On Image Dehazing Based On Fusion Of Multiple-depth-layers

In recent years, haze weather has become a frequent phenomenon which severely hinders traffic safety and travel convenience. Images received by sensors under hazy conditions are drastically degraded due to the atmospheric scattering of light. Concealed by the atmospheric veil, objects are difficult to identify. Such degradation results in images with low contrast, dim colors and fuzzy scenes. The dehazing process, which makes images or videos more visible and clear, can improve the quality of data for furthur manual or machine interpretation, thus has important theoretical significance and practical value.This paper first introduced the physical model of haze imaging, followed by several relavant prior theories, and analyzed the features of hazy images. Based on these theories, a hazy image detection method was proposed, which could automatically judge whether the input image is hazy by observing color and intensity statistics. Then a novel dehazing approach that combined the imaging model with image processing methods was studied. This method employed a strategy of image fusion instead of direct parameter estimation. A sequence of processed image layers was first derived with different depth parameters, which are extracted from the input image, and was then blended into a single result according to qualitative assessments of each pixel of each layer.This paper tested the accuracy of proposed haze detection method, validated proposed dehazing method by experiments and evaluated its subjective and objective performance. Results showed that the proposed hazy image detection method made right judgement most of the time, and could roughly indicate the influence of haze. The proposed dehazing mothed could evidently improve the visibility of hazy images, yielding bright, vivid and clear results.