Extraction of objects contours for body language study

Abstract

A new method for contours extraction in grayscale images for body language study is presented in this paper. The image processing is based on two kinds of image filtration, used to suppress the image textures and to smooth the image background, retaining the main contours. For the textures suppression, we use an adaptive fuzzy filter and for the background processing we use a 2D linear digital filter of a non-recursive kind. The contours extraction is done with image segmentation based on the “Triangle algorithm”, specially modified for this application. The intelligent contours extraction is very useful for multiple applications, such as body language understanding, sign-language representations for training of deaf people, etc. The contour images permit very efficient lossless compression with high compression ratio, which is a base for the development of various application tools for the mobile communications and distance learning.