Research On Foggy Image Restoration Based On Physical Model

In the foggy weather, both the radiation effect of the particles to the scene in the atmosphere and the scattering effect of the airlight cause poor visibility, leading to the contrast and color saturation of images captured by a outdoor vision system degrade, which has an influence on the observers’ visual feeling and the accurate extraction of useful information. Therefore, it has very important theoretical significance and broad application prospects to restore degraded image affected by fog. Currently, the previous algorithms have many limitations, thus image quality can still be further improved. This paper analyzes the physical model and degradation mechanism of the fog imaging, and explores key technology and implementation for image defogging. The main works of this paper are shown as follows:1) Based on the physical model, the refining and edge-preserving properties through guided filter, bilateral filter and recursive filter are respectively used to estimate the atmospheric veil. The calculation method of the airlight is improved by computing the average values of the most dense fog. The visual effect is further improved through enhancing or weakening local regions. Experimental results show that the proposed algorithms combining multiple non-linear smoothing filters have very high efficiency and achieve great restoration for color information and detail feature.2) According to the existing methods have the defect of insufficient scene adaptability, a defogging algorithm combining image fusion and segmentation is proposed. The morphology operation and weighted image fusion are adopted to estimate the fog density rapidly. Segmentation thresholds are set according to the overall brightness and the local variance of the image, thus can be used to positioning the densest foggy region. Both the subjective assessment and the objective indicator demonstrate that the proposed method has robust scene adaptability and achieves different degrees of improvement in terms of restoration effect and computation speed.3) Since the image after defogging has problem with poor chroma and low brightness, an effective solution using tone mapping based on human visual system is proposed. A simple function fitting from Web’s law is adopted to adjust according to the fog density of scene. The recovered image quality is further enhanced at a low computing cost.4) In view of the road scenes and remote sensing images need to clear in the engineering applications, this paper analyzes and enumerates the basic essential features from both special scenes, and obtains the accurate atmospheric veil combining Gaussian filter. In addition, a simple but efficient estimation of airlight is proposed according to these features. Experiments performed in both application scenes demonstrate that the new methods effectively remove the influence of fog and greatly improve the image visibility.