The Matching Method for Rectified Stereo Images Based on Minimal Element Distance and RGB Component Analysis

摘要

A common problem occurring in medical practice is the localization of veins and arteries. To determine the location of these elements, it is not necessary to have a complete 3D model. A much better solution is preliminary segmentation yielding the contour of veins, and further search for stereo correspondence already in binary images. The computational complexity of this approach is much smaller, which guarantees its fast operation. The disparity matrix is created according to the principle that the most likely correct distance between the same elements in the left and right images is the minimum value. Then, the adjacent RGB components surrounding the elements aspiring to be homologous are analysed. The operation of the method is illustrated on the basis of the authors' own images as well as standardized images. In addition, its operation was compared with three recognized and widely used algorithms for image matching. The effectiveness of the new method reaches less than 94 % of correctly matched pixels with a standard deviation of 1.5 pixels and operation time of 90 ms.