Rotation and Scale Invariant Color Image Retrieval Using Fuzzy Clustering.
Shan Li, M. C. Lee · 2005
Color is one of the most important image features used in content-based image retrieval (CBIR). This paper proposes a robust and effective color feature known as ring-based fuzzy histogram (RFH), which reduces noise sensitivity in conventional color histograms (CCH) by using two stages of fuzzy clustering. By partitioning images into rings, RFH can capture an image’s spatial information, and is robust to rotation and scale variations. This paper also presents details about the extraction of RFH. The fuzzy histogram extracted from the whole image is used to filter out highly dissimilar images during the image retrieval process, leading to a reduced query response time. Experiments results show that the proposed feature is, in general, compact; robust to changes due to images rotation or image scaling; and robust to illumination changes and quantization noise. It also outperforms CCH and other recently published proposal, such as fuzzy color histogram [6]. 1.