Robust Image Retrieval Approach Based on Clustering Analysis of Re-Quantized Color Image

Abstract

In this paper, we propose a robust approach about color image retrieval. It can realize fast matching in CBIR (Content-Based Image Retrieval) when we search in large image databases. Indexes root in object features of Z image which is the result of re-quantization in HSV color space, matching with a non-geometrical distance is based on objects, so time consumption pixel by pixel can be avoided. Because Z image is made up of many color clustering regions and invariant moments are used for feature representation, our approach is robust to translation, scale and rotation.

Fast Clustering-Based Kernel Foley–sammon Transform

“…20-22 used automatic subspace clustering for high dimensional data, Ref. 23 used clustering analysis in robust image retrival, and Ref. 24 advanced an efficient and effective clustering methods for spatial data.…”