Arbitrary background picture segmentation
Samuel Itzikowitz, S. Sheraizin · 2003
In this paper, we present a new method for real time arbitrary background picture segmentation. We consider the following picture conditions: color picture, noisy picture, scene light changes, and still image arbitrary background. They are typical for many applications, e.g. for video security system, videophone, videoconference, V-commerce, etc. A set-theoretic approach has been used for picture model creation, adaptive picture processing and noise reduction. The presented method main problem, namely noise immunity and resolving power, is investigated by using picture clustering. The both luminance and chrominance picture components use allows to avoid an influence of "undesirable" object light changes and noise. It is shown that for arbitrary background picture segmentation, the required signal-to-noise ratio must be from 40 dB to 44 dB, depending on the frames change rate.