Improved Techniques for Mammographic Image Compression using Balanced Multiwavelet Block Tree Coding

The recent innovation in digital medical imaging techniques requires the development of high-performance storage and image transmission systems. The biggest problem in digital technology is voluminous amount of data generated. For example, a digital mammographic image of size 1024x1024 pixels with 8 bits per pixel requires 8.3 Mb for storing it in original form. Researchers in several studies have demonstrated the need for high rate compression algorithms for medical imaging applications. But the recently reported image compression results indicate that Peak Signal-to-Noise Ratio (PSNR) is outperformed by scalar wavelets. However, it often fails to capture high-frequency information accurately. Interestingly, multiwavelet preserves high-frequency information in mammographic image and provides good energy compaction. The challenge still remains as to how one can better represent the signal for achieving the best compression. This paper proposes a solution to the above open problems using balanced multiwavelet-based image compression. The proposed scheme presents combination of two novel ideas: A coefficient reorganization suitable to balanced multiwavelet decomposition is used to regain the parent-child relationship. A block tree coding is used for compression and reconstruction of multiwavelet transformed image and thus the proposed scheme is called as Multiwavelet Block Tree Coding (MBTC). This approach yields the advantages of high energy compaction, PSNR, and a less number of bits for encoding. Balanced multiwavelet-based compression, with MBTC applied to a set of four mammographic images, achieves an average PSNR of 43.245 dB against the existing Set Partitioning In Hierarchical Trees (SPIHT) algorithm which on an average achieves only 34.181 dB for 0.5 bpp bit rate and MBTC requires encoding bits of 45.565 less than SPIHT encoding.