A method of leg skin recognition based on distribution of skin texture
Peng Yao, Han Su · 2017
Identifying the criminal and victims in images with only parts of skin existing is a new and challenging task. For this situation, the individual recognition in tradition is invalid, because there are not obvious features existing in the skin images, especially in some forensic cases, there are neither faces nor body labels can be observed. To address this problem, some methods based on skin mark pattern and blood vessel pattern are proposed, however, these methods neglected a fact that the image is not always high resolution, skin marks and blood vessel are not reliable sometimes. A recent paper indicated that androgenic hair patterns are effective in low resolution, but the alignment is not considered and the matching is not robust to viewpoint changing. In this paper, a new feature pattern based on distribution of skin texture trend which firstly borrows the conception of contour line in geography is proposed. A sliding block system is designed to increase discrimination ability and robustness to rotation. Different legs with resolution of 25, 18.75, 12.5, 6.25, 2.60 and 1.30 dpi were examined. Experiment results demonstrate that the proposed algorithm is effective and has rotation invariance, which has a certain improvement.