Radiological image compression using wavelet transform implemented on inexpensive hardware

Digital radiographs require large storage space. Transmission of the radiographs in applications such as Picture Archiving and Communication System (PACS) and teleradiology requires large bandwidth. The bandwidth problem is especially severe when the radiographs must be transmitted over slow communication lines such as telephone system or ISDN lines.; Reversible and irreversible image compression methods are developed to reduce the storage and bandwidth requirement of digital radiographs. The methods are based on discrete wavelet transform (DWT) implemented using two-channel filter banks. The filter banks are designed to perform signal expansion on biorthogonal discrete sequence bases. In reversible compression, the filter banks also have rational coefficients and the expansion coefficients are directly encoded using arithmetic coding with an order-0 finite context model. In irreversible compression, the expansion coefficients are quantized using either subband noise minimization technique or subband perception sensitivity technique. The quantized coefficients are then encoded with the same arithmetic coder as in reversible case.; The irreversible compression performance is compared to methods based on full-frame discrete cosine transform (FFDCT) and the method based on JPEG standard. It is found that the new methods performs better than FFDCT methods in terms of normalized mean square error (NMSE) and visual quality. The new methods performance perform comparably to JPEG both in NMSE and visual quality. The reversible compression can achieve 2:1 to 4:1 compression ratio routinely.; The possibility of incremental transmission of radiographs, successively adding more details, is also investigated. It is found that, for reversible compression, 30% of compressed data can be transmitted to reconstruct a good approximation of the full reconstructed image. For irreversible compression, 50% of compressed data must be used for reconstruction to achieve comparable image quality.; The methods can be implemented using commonly available microcomputers without specialized hardware running widely available softwares.