Optimized architecture for trace transform based CBIR system

The rapid growth of internet technology combined with affordable storage capabilities has lead to many challenges for efficient retrieval of information especially images for real time applications. One way of supplementing the software approaches is to provide hardware support to the system architecture. We propose an algorithm and architecture to implement trace transform based CBIR system. The proposed algorithm is optimized for time and space parameters by reducing the addressing complexity of the feature extraction phase. Further optimization is achieved in the similarity phase by using hamming distance measure and threshold based sorting algorithm, which is possible because the identifier is a binary string. Synthesis results show that the above optimizations have led to reduction in the response time of the retrieval process and we are able to process 8389 images per second. A speed up factor of 1079 is achieved on the hardware platform when compared to the implementation of the proposed algorithm on the software platform.