1-D DCT shape feature for Image Retrieval

Abstract

Content Based Image Retrieval (CBIR) is a difficult task. It needs different approaches and tools to get a set of images similar to a complex query image. Finding appropriate features is one of the important issues in this process. There are several tools to extract the appropriate features from the images. In this paper we explore the usage of 1-D Discrete Cosine Transform (DCT) as a tool for finding features of the boundary of objects in the query and database images. It is shown that the sensitivity of the DCT to rotation of objects can be reduced and the 1-D DCT of extended boundary signals can even give better results than the DFT in shape retrieval process.