Research And Implement Of Medical Imaging Texture Analysis Based On Two Dimensional S Transform

The two dimensional S transform can be used in medical imaging processing for its capacity of multiple dimensional analysis in arbitrary frequency. The huge time and memory consumption of the implementation about the two dimensional S transform have been the largest obstacle of its application in medical imaging processing. We can reduce the time and memory consumption by the implementation of orthogonal method, but the segmentation of the Flourier frequency will involved in the ring artifacts which will reduce the quality of the medical image during the implementation of the orthogonal method. To overcome these defects, we advance a new method based on the oversampling to implement the two orthogonal dimensional S transform according to the overview of multiple dimension analysis at home and abroad, and apply it into the segmentation of small anatomical structures in lung medical imaging processing. The main achievements are summarized as following.1) In this thesis we have overviewed a variety of multi-scale analysis method, and advance the fast implement of two dimensional orthogonal S transform based on oversampling method on the basement of detailed analysis about the character of the two dimensional S transform. The algorithm combined the advantages of the oversampling method with the orthogonal S transform which can successfully reduce the time and memory consumption can effectively reduce the ring artifacts introduced by the implementation of orthogonal two dimensional S transform at the same time.2) We design and implement the two dimensional orthogonal S transform based on oversampling method in the IDL platform, and compare the time and memory consuming with the two dimensional S transform of arbitrary dimension and format.3) We introduced the orthogonal two dimensional S transform based on oversampling into the segmentation of small anatomical structures in lung CT medical imaging processing. The result of the experiment demonstrates that the method we advanced can more effectively segment these tissues by determining a reasonable energy threshold than the traditional threshold approach under the present experimental conditions, such as the segmentation of capillary, cleft between the linear levels of pulmonary fissure.