Main Feature Extraction and Expression for Religious Portrait Thangka Image

摘要

A method based on the least-squares curve fitting of orthogonal multinomial to express image contour characteristic was proposed in this paper, which obtained successful application in head ornaments feature extraction and expression of Buddha, Bodhisattva, Spiritual teacher of Thangka portrait image. First, according to image components' graylevel histogram, chooses an appropriate threshold value for further thresholding; Second, transform pixel point of the image to coordinate points under rectangular coordinate system, then extract typical contour and eliminate the imaginary point. Third, the contour point which fits a multinomial with the leastsquares orthogonal multinomial distance fitting method, compares three data with its square coefficient and four powers coefficient as well as goal object's proportion of high and width of Thangka image to express the image characteristic as feature vector. Experiments show that the method is simple and effective.