Partially lossless compression of DICOM image sets

Abstract

In this paper a new approach to the compression of DICOM image sets is presented. For this task, a specific image projection space is created by employing principal component analysis. A subset from the initial image set, which can include regions of interest, is selected and used in the projection space creation. That way, these images can be reconstructed lossless, while the residual images are reconstructed with little loss of information. As shown by conducted experiments, the proposed method yields mean absolute errors below those of linear interpolation methods, yet at the same time achieves evidently higher compression ratios than compared image compression algorithms.