An Approach to Partition-Based Color Image Retrieval
Guoyu Wang · Qingdao Haiyang Daxue xuebao · 2005
The traditional color histogram contains the component of color but lacks spatial information. To improve the performance of content-based image retrieval, spatial information is indispensable. An approach to partition-based color image retrieval is proposed in this paper. The images in the training set are partitioned into size fixed cells and, from each cell, the local color histogram as a color invariant feature is extracted. By clustering and coding, the images in the database can be regarded as a combination of local color features in certain spatial distribution. The effectiveness of this approach is comfirmed by the experimental results.