A Research Of Image Enhancement Algorithm Based On Retinex Theory

Image enhancement is one of the methods which is commonly used in digital image processing. Image enhancement is the processing method which highlights some of the information in a specific image based on the need, and weaken or remove some unwanted information. The main purpose of image enhancement is to make the processed image of a particular application more suitable than the original image, and the image processing. result more suitable for observation of the human eyes or machine recognition system. It can be applied to some fields such as pattern recognition, medical images, remote sensing images, digital image processing and computer vision and so.The feature of traditional image enhancement algorithm of two kinds of spatial and frequency is single.The image enhancement algorithm can only enhance a kind of image or some features of image. The enhancement algorithm based on Retinex theory has the advantages of sharpening, color constancy, dynamic range compression, and high color fidelity. This thesis is try to give a comprehensive analysis of Retinex enhancement algorithm,what is more,to do some theoretical research and to improve and implement the McCann algorithm.The main task as follows:(1) Proposed a new initial value estimation method. For large differences in brightness of color images, the blocky effect emerges because of the initial estimation of non-smooth value resulting. For this fault, this paper uses the largest sub-regional average to replace the global mean method as the initial image estimate to amend to enhance the image, to improve image brightness and contrast, and to get great visual effect.(2) Proposed the new comparison path of light and dark. The new path can show the changes of illumination in image and can not weaken the effect of McCann enhancement algorithm. When the brightness is estimated, it can improve the efficiency of the algorithm by reducing the number of comparison pixels. (3) Improved the algorithm of color image enhancement based on McCann algorithm. In this paper, color image is processed in HSI space. The image is stretched by using two different methods of stretching to get enhancement image, and then get the final image by fusing enhanced image. It not only improves the image brightness, but also reduces the color distortion.(4) Based on the improvement of the proposed method,multiple grayscale images and color images are simulated. The enhancement experiments of Gray-scale image shows that the brightness and the information entropy of the image are improved and the enhancement effect is ideal in the local region of the image. The algorithm gets enhancement images with a large dynamic range compression. The enhancement experiments of color image shows that the algorithm has more good color constancy, and has high color fidelity and brightness.Finally, there is to give a simple analysis of the results of enhance image, to get a valid conclusions of the improved algorithm in this paper...