S. Anishenko (Russia, UK), D. Shaposhnikov (Russia), R. Comley, and X. Gao (UK)
Colour appearance model, video image processing, CIECAM, face segmentation, motion detection
For tracking head motions from a sequence of video images, segmentation of faces usually takes place first to locate head positions. Although a plethora of methods have been developed to segment faces, each approach has its own advantages and disadvantages depending on the properties of collected video clips, especially under low luminance level (i.e. lower than 10 cd/m2) . This study extends a traditional approach based on skin colour to segment faces, and has arrived at a mixed-colour-space algorithm, which takes viewing illumination and post position into consideration and shows very promising segmentation results with fewer false regions. The procedure includes obtaining colour attributes that are selected from varying colour spaces/models in order to detect the regions of interest (ROIs), which maximises the characteristics of faces. Colour edges are then applied to classify ROI into faces.
Important Links:
Go Back
