| 摘要 |
The present invention relates to a depth information based human skin region detection method for labeling pixels having similar depth features from a stereo three-dimensional image and detecting regions, having human skin color distribution among labeled regions, as a real skin color region, the method comprising the steps of: (a) extracting depth information from the stereo three-dimensional image; (b) performing labeling with respect to images of the depth information; (c) detecting a skin color region from the three-dimensional image; (d) detecting a candidate group of a skin region by performing AND combination between the labeling region and the skin color region; and (e) detecting a final skin region by evaluating texture smoothness in the candidate group of the skin region. According to the skin region detection method as described above, by effectively combining color features and depth features, errors of skin color regions which are incorrectly detected in an existing background region can be significantly removed, and, as a result, a skin region in general can be more exactly extracted. [Reference numerals] (AA) Start;(BB) End;(S10) Extract depth information from a three-dimensional image;(S20) Perform labeling with respect to images of the depth information;(S30) Detect a skin color region;(S40) Detect a candidate group of a skin region by performing AND combination between the labeled region and the skin color region;(S50) Detect a final skin region by using texture smoothness |
