Study on Object Contour Extraction Based on Hölder Exponent and Multifractal Spectrum

Abstract

Object contour extraction is one of the most important steps for object segmentation and recognition. The edge detection method based on gradient detection have some disadvantages if there are some noise in image, the edge errors occur in the gradient detection stage will be transmitted to the edge connection stage and the errors can’t be modified during the procedure. Multifractal spectrum gives a complete statistical description of the singularity in the image. In this paper, a new method based on the combination of the singularity and multifractal analysis is proposed to extract object contour. The experiments show that the contour extraction results are encouraging.