Skin area detection for video image systems

摘要

Apparatus for detecting skin areas in video sequences includes a shape locator and a tone detector (both in 12). The shape locator analyzes the input video sequences (26) to identify the edges of all the objects in a video frame and determine whether such edges approximate the outline of a predetermined shape that is likely to contain a skin area. Once objects likely to contain skin areas are located by the shape locator, the tone detector examines the picture elements (pixels) of each located object to determine if such pixels have signal energies that are characteristic of skin areas. The tone detector then samples pixels that have signal energies which are characteristic of skin areas to determine a range of skin tones and compares the range of sampled skin tones with the tones in the entire frame to find all matching skin tones. An eyes-nose-mouth (ENM) region detector is optionally incorporated between the shape locator and the tone detector to identify the location of an ENM region on an object that is likely to be a face, so as to improve the accuracy of the range of skin tones that are sampled by the tone detector. <IMAGE>