Segmentation using multifractals for object-oriented video coding

Abstract

This paper presents a study of the application of multifractal measures in the segmentation of a video sequence image for object-oriented video coding through the generalized Renyi entropy. Grey-level sequence images are analyzed from the point of view of strange attractors. Multifractal feature maps are used to extract the features from the video sequence image. The nonuniform property of the image is reflected in the singularity spectrum and the Mandelbrot spectrum. Each video-object plane (VOP) specifies a particular image sequence content and the textures in the different contents can be separated because similar textures generally will have an homogeneous property which can be characterized by the singularity and Mandelbrot dimension of the fractal sets. Consequently, we can achieve not only image segmentation, but also a separation of the video-object layers (VOLs).