OBJECTIVE METHODS FOR OPTIMIZING JPEG COMPRESSION OF CORONARY ANGIOGRAPHIC IMAGES

Digital angiographic images contain a significant amount of redundancy as well as some irrelevant information and noise. Therefore, it is possible to reduce the number of bits required to represent an image considerably. The lossy JPEG standard may be used provided that no significant diagnostic information is lost. As implemented in presently available hard- and software in most cases the luminance quantization table (LQT) is applied for gray level images, which may only be scaled by a so-called quality factor. The questions arise whether it is possible and worthwhile to specify quantization tables for the particular characteristics of angiograms. To assess the quality performance quantitatively, global numerical quality measures and evaluations based on Hosaka-plots were performed. Those diagrams compare the errors introduced into areas of different local activity. By the newly introduced weighting of these errors with the relative occupancy of the respective classes of activity the results got more reproducible. The blocking and blurring effects introduced by lossy JPEG compression could be compared objectively. Two new quantization tables were derived from the transfer function of the angiographic X-ray system, the modulation transfer quantization table (MTQT) and the star pattern quantization table (SPQT). Both tables guarantee that the blurring of sharp edges is minimized so that no deterioration around a coronary lesion occurs. Based on the signal-to-noise ratio, the overall quality performance is the same as for the LQT. A general relation between the bit rate of the compressed image and the quality factor has been determined for images of high local activity and normally scaled coronary angiographic images (512 x 512).