An Improved Chromatic Skin Color Model for Detecting Human Skin in JPEG Images

The detection of the human skin has proven to be a useful and robust technique in computer vision for detecting, segmenting, and tracking faces and hands. It is also used in many different applications in motion capture, human-computer interaction, access control, surveillance, and content-based image retrieval and indexing of image databases. To build a decision rule that will discriminate between skin and non-skin pixels, a metric has to be introduced to measure the distance of the pixel color to skin tone. The skin color models define the type of this metric.An improved chromatic skin color model is discussed in this thesis. The model detected the human skin in the Bitmap images with JPEG format. A threshold method and 2D Gaussian model were used to improve the accuracy of the skin region detected. Good results have been achieved by using the proposed model. Results show that the model detects the human skin with an excellent detection rates for over than 90%. This result shows that this model is more appropriate to avoid false detection areas, while there remains a high degree of correct detection. This model has been implemented using the MATLAB software, which able to improve the performance and accuracy of skin detection models.

Text FK_2005_59.pdf Download (1MB)

Actions (login required)

View Item