Objects Separated From X-ray Images Based On CS And ICA

The X-ray medical imagesis of referential value for it can be used to examine the diseased tissues of the patients. But the noise, the artifacts and the overlapped organs always damage the X-ray image information. In this paper,based on Independent Component Analysis(ICA) and Compressed Sensing(CS), a method of separating objects from the multi-energy images is proposed.ICA is used to separate the objects from their multi-energy images; moreover,CS is used to improve this processing in both runtime and performance. From the multi-energy x-ray images, a set of x-ray images is selected as the ICA observation vector, then the vector is sparse transformed based on some kind of sparse basis. After ICA is used to separate the sparse vectors, the separated sparse vectors are inversely transformed into the spatial domain to be the reconstructed imagesby using Orthogonal Matching Pursuit(OMP).Conclusions are that, based on ICA and CS, the overlapped objects are separated successfully from the x-ray images which are produced with three different energy; compared with using ICA only to separate, using CS and ICA improves the processing performance in decreasing runtime by 46.14s(23.3%)and memory occupancy by 21%, while increasing Peak Signal to Noise Ratio(PSNR) by 2.665 dB, edge gradient by 0.001 and information entropy by 0.09. Conclusion: The objects can be separated successfully from the multi-energy images using ICA, which can be improved further by using ICA combined with CS.