Image back-light compensation with fuzzy C-means learning algorithm and fuzzy inferencing

In this paper, a two-stage processing technique utilizing the fuzzy c-means learning mechanism and the fuzzy logic rule inference is proposed to compensate the back-light images. The advantages of this approach are: (1) the subject region can be compensated independently, (2) the brightness of subject region can be enhanced without interference of the background, (3) the contrast of subject region can be enhanced adequately, (4) the original background image can be reserved. We have implemented the proposed compensation method and tested on 30 pictures. The performance is better than that of global histogram equalization and global brightness adjustment.