Multi-scale color image enhancement algorithm based on fuzzy peer groups

The advantages of different single scale Retinex algorithms in detail enhancement and chromaticity preservation can not be utilized sufficiently by existing multi-scale Retinex algorithm due to using fixed weighted value. Moreover, the noisy pixels are often enhanced in the enhancement process. Aiming at these disadvantages, we propose an improved algorithm based on fuzzy peer group. First, the pixels are classified into noisy ones, detail ones and smooth ones using fuzzy peer group technique and the noisy pixels are suppressed with the vector median filter. Then the image is enhanced with different scale Retinex algorithms and the results are fused according to the proportions occupied by different pixels in local window. Finally, further operations of color restoration, dynamic truncation and stretch are carried out to solve the problems of color distortion and narrow dynamic range. Experiment results show that the proposed algorithm outperforms other methods, and can be applied in engineering application and has practical application values.