Shape Feature Extraction Using Fourier Descriptors with Brightness in Content-Based Medical Image Retrieval

Abstract

Contour-based shape feature extraction is one of the important research contents in content-based medical image retrieval. The paper presents a method using Fourier descriptors with brightness. The method uses centroid distance function to compute shape signature from boundary pixels of a shape. Fourier transform is used for shape signature to compute Fourier coefficients, and standardized pixel brightness is introduced into computational process of the Fourier coefficients. The Fourier coefficients which are invariant to translation, scaling, rotation and change of start point are used as Fourier descriptors. And shape feature vector consists of the Fourier descriptors. Experiments show that the system which uses the method in the paper has better performance than that which used Fourier descriptions in terms of overall performance.