Research Of Clearness Theory And Methods For Fog-Degraded Images

In bad weather, such as fog and haze, the brightness and color of images acquired by imaging device are affected by the scattering particles in the atmosphere. This results in severely degrading of image quality and great reduction of the application value of videos and images. A fast and efficient algorithm for image clarification is proposed and implemented based on in-depth study for image enhancement and restoration methods which are respectively using the increase in contrast and fog degradation mechanism. The proposed research results are evaluated by subjective observation and objective data analysis. The main contents are as follows:1) This dissertation analyzed and discussed the research background of image enhancement and image restoration in detail. The basic principles of several typical hazy image clarification methods were expounded, and the advantages and disadvantages of these algorithms were analyzed.2) For the deficiency of the Automatic Color Equalization algorithm can not keep the original color information well when using for hazy image enhancement, a double selection pixel of local adaptive ACE algorithm was proposed. The proposed algorithm transformed the obtained images from the RGB color space to the YCbCr color space firstly, and then employed gradient information of the luminance image and local information of the image to select pixels in double layer. Lastly, it adjusted the brightness control function in ACE algorithm with locally adaptive filter. The algorithm is not only good at color retention in nature, but also calculating fast.3) For the high complexity of medium transmission computing, He’s dark channel prior knowledge can not satisfy the real-time requirement when enhancing hazy image. A new high precision and fast computing technology is proposed for estimating transmission and haze-removal in this dissertation. First of all, this paper appropriately improved mean shift algorithm, and then precisely figured the transmission map obtained by dark channel prior algorithm through improved mean shift. The proposed algorithm was introduced into simple image haze remove using dark channel. The algorithm eliminates the block diagram and halo phenomenon of the transmission map, and solves the problem of rapid calculation for the high precision transmission.4) This paper implemented experiments on some outdoor foggy images. The results show that the improved algorithms get better visual effect, keep the original color when restoring image detail and ensure the efficiency compared with the conventional classical fog clearness methods.