The feature extraction and similarity measure of CBIR algorithms involve highly computation intensive and repetitive operations on a large data set yet have to satisfy the real time application requirements. One way to supplement software approaches for this purpose is to provide hardware support to the system architecture. We propose an algorithm and architecture for hardware implementation of trace transform based feature extraction for CBIR system. The proposed algorithm is focused to reduce the computational complexity in the addressing block for trace based feature extraction and is also optimized for memory consumption and speed for distance calculations in the similarity measure phase. Synthesis results show that the above measures are responsible for reducing the response time of the retrieval process by being able to process 2725 images per sec. © 2011 Springer-Verlag.