Automated segmentation of regions of interest from thermal images of hands

Thermal imaging can provide an image of the surface temperature of an object in a non-contact and noninvasive manner, making it particularly appealing for use in medical applications. In applications where it is desirable to extract temperature data from anatomical regions of interest (ROIs) in a standardised and consistent manner, the use of automated segmentation and analysis techniques can provide a faster, more reliable and more consistent approach than manual segmentation of these ROIs. In this paper we present an algorithm which automatically extracts temperature data from eight ROIs in thermal images of the volar aspect of human hands. The algorithm first identifies the hand from the background in the thermal image and then identifies pixels which make up the fingers and the palm. Finally, eight ROIs are extracted from the identified regions. The methods proposed in this work can also be extended for the processing of similar visual images.