Application of Gray Level Variation Statistic in Gastroscopic Image Retrieval

Content-based medical image retrieval is getting more and more importance in aspect of clinical assistant diagnose. In this paper a system for gastroscopic image retrieval is developed which is available to support clinical decision making. First a new method based on texture feature is proposed which statistic the gray level variation of each pixel in 3x3 domain. And then Earth Mover's Distance is used to calculate the dissimilarity. Meanwhile, a method combining both color and texture is proposed to carry out integrate retrieval. Finally, some contrast experiments are designed in the retrieval accuracies, the rank and the execution time. The comparison of the experimental results shows that the approach proposed in this paper is effective.

1. Introduction

With the rapid development of modern imaging technology and information processing technology, the research of content-based medical image retrieval (CBMIR) has got more and more attentions [1]–[3]. Because of medical image's inherent characteristic, the retrieval methods in content-based image retrieval (CBIR) can not be simply repotted in CBMIR. The key issue is how to retrieve the right ones that could be used in clinical diagnosis, curing and the aspect of studying in the database.