DR Image and Fractal Correlogram: A New Image Feature Representation Based on Fractal Codes and Its Application to Image Retrieval

摘要

In this paper, we propose a novel image feature representation method using fractal codes and its application to image retrieval. A fractal code consists of contractive affine mappings; each mapping represents a similarity relation between two regions in an image. A fractal-coded image is composed of a set of fractal codes. In the encoding process, the region sizes in the fractal codes tend to reflect the local features in the original image. We propose a method to represent this size as a feature image termed as a 'DR image'. We also propose an efficient retrieval method using correlograms of DR images. This retrieval method is an application of DR images and can be used in the compressed domain. The effectiveness of the proposed methods have been illustrated by several experiments using an image database released on the Internet.