Based On Retinex Image Enhancement Algorithm Realization

The Retinex theory is based on adjusting human visual system and perceiving object color and luminance model, which is proposed as image intensification theory. Its essential ideology is that we estimate illuminative image from original image, then we subtract the illumination image with the original image. Thus, we can obtain reflection image that can response the essence of objects. And the reflection image is the nature appearance of the objects in images. Retinex image enhancement methods can effectively compressed the dynamic range of images, to make image detail in dim area more prominent, enhance contrast of image and maintain effectively image color and so on. Therefore, Retinex theory has a more widespread application in the image enhancement domain. In this paper, the theory would also be focused.This thesis first have implemented several kinds of classical iterate-based and surround-based image enhancement method, while studying these Retinex image enhance method which is based on the color constant theory, Then measured the advantage and disadvantage of enhanced image reasonable using common indicators of image quality evaluation by experiment. This thesis also evaluated advantage and disadvantage of these classical methods in improving image quality, maintaining image color and computing speed while analyzed the common good and bad points of these method, and advantage and disadvantage of different indicators to their respective different method.Finally, the thesis gave detailed description and simulation to the classical Retinex method. When it makes image enhanced, it lead to produce halo. Then one method to eliminate halo phenomenon have been proposed. With analysised and compared the experimental results, the experiment have proven that the new Retinex method can not only improve image quality, but also reduce halo artifacts.