Image indexing using the color and bit pattern feature fusion

• Color image indexing is proposed by employing the low complex ODBTC for generating features.
• Color co-occurrence feature from the ODBTC quantizers captures the color distribution.
• Bit Pattern Feature (BPF) characterizes image edges and visual patterns.
• Experimental results demonstrate that the proposed method is superior to former schemes.

This paper presents a new way to index a color image by exploiting the low complexity of the Ordered-Dither Block Truncation Coding (ODBTC) for generating the image features. Image content descriptor is directly constructed from two ODBTC quantizers and the corresponding bitmap image without performing the decoding process. The color co-occurrence feature (CCF) derived from the ODBTC quantizers captures the color distribution and image contrast in block based manner, while the Bit Pattern Feature (BPF) characterizes image edges and visual patterns. The similarity between two images can be easily determined based on their CCF and BPF under a specific distance metric measurement. A metaheuristic algorithm, namely Particle Swarm Optimization (PSO), is employed to find the optimum similarity constants and improve the retrieval accuracy. Experimental results demonstrate that the proposed indexing method is superior to the former Block Truncation Coding (BTC) image retrieval system and the other existing methods. The ODBTC method offers an effective way to index an image in a content-based image retrieval system, and simultaneously it is able to compress an image efficiently. Thus, this system can be a very competitive candidate in image retrieval applications.