An Improved Homomorphic Filtering Algorithm for Face Image Preprocessing

An Improved Homomorphic Filtering Algorithm for Face Image Preprocessing

Homomorphic filtering has a good illumination preprocessing effect on face images under complex illumination conditions such as weak illumination, side illumination and drastic illumination changes. However, too many parameters need to be set artificially in the filtering process, and the pre-processed face features are not clear. In view of the above disadvantages, Firstly, this paper proposes an exponential homomorphic filtering image enhancement algorithm with only one parameter. In order to highlight the face details after illumination pre-processed, this paper improves the traditional canny operator: adding 45° and 135° gradient directions of the edge template, and then using the improved canny operator to extract the edge of the face image after the improved homomorphic filtering, so as to get a more complete face edge image. Finally, this paper weighted fused the edge image and the face image after illumination pre-processed, and a new approach of homomorphic filtering for illumination pre-processed is proposed. Simulation results show that the details of face images pre-processed by this method are clear and the overall contrast is moderate. Compared with the traditional algorithm, the accuracy of face recognition is improved by almost 2% in the extended Yale B face database, which proves that the proposed method is effective.