High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

Skin Segmentation in Images and Video based on Stacked Auto Encoders (SAEs)

Author(s):

Berlin S John , MOUNT ZION COLLEGE OF ENGINEERING, KADAMMANITTA, PATHANAMTHITTA; Sudheesh S R, MOUNT ZION COLLEGE OF ENGINEERING, KADAMMANITTA, PATHANAMTHITTA; Anitha Rajam B S, MOUNT ZION COLLEGE OF ENGINEERING, KADAMMANITTA, PATHANAMTHITTA

Keywords:

Pixels, SAEs, Skin Segmentation, Encoders, Images

Abstract

Traditional skin segmentation is based on skin probability maps and guassian mixture models to capture skin tones under different conditions . A good skin detector is important for human-machine interaction applications. The greatest challenge in the traditional skin segmentation is false positive rate and overlapping of skin pixels and non- skin pixels. This paper presents a novel skin feature learning algorithm based on Stacked Auto Encoders(SAEs). Instead of predicting each pixels individually, we utilize block of pixels for skin segmentation to avoid overlapping of skin pixels and non-skin pixels. The algorithm exploits the learning ability of deep neural networks to learn high-level representations of skin tones.

Other Details

Paper ID: IJSRDV5I120531
Published in: Volume : 5, Issue : 12
Publication Date: 01/03/2018
Page(s): 990-991

Article Preview

Download Article