基于Retinex理论的图像增强算法*

这种增强算法首先依据像素的R，G，B分量将输入的彩色图像被分解成为三幅图像，代表场景中波长不同（长波、中波和短波）的反射光的强度；分别计算长波、中波和短波波段内像素间的相对明暗关系，进而确定每个像素的色彩。最后，将Retinex色度空间内的色彩线性映射到RGB空间，获得经过增强的图像。通过这种方法所获得的图像具有色彩逼真度、动态范围大的特点。计算机仿真结果表明运用这种方法进行图像处理可以获得非常好的处理效果。

Image Enhancement Algorithm Based on Retinex Theory

Input color image is first divided into three pictures according to its R, G and B values, and each one represents the intensity of reflected light with long-, medium-and short-wavelength. In order to determine the color of each pixel, these three pictures are then processed independently to calculate its lightness. Finally, color in Retinex color space is mapped to normal RGB space to reconstruct enhanced image. Study shows the resulting image often has better color fidelity and larger dynamic range. Computer simulation results demonstrate that it is possible to achieve very good performance by making use of this theory correctly.