Segmentation of Single Image

First, a typical carotid MR image is shown in Fig. 8.10(a). Because of the noise or artifacts during imaging process, the intensity of lumen area is not uniform and there are some isolated bright spots inside. The QHCF algorithm was applied to this image and segmented it into many regions as shown in Fig. 8.10(c). From the result, we can see that the lumen segmentation is not affected by those bright spots inside the lumen area and most of the noise in the background have been suppressed. This is better than the result segmented with adaptive ICM algorithm shown in Fig. 8.10(b). By tracking the boundary of lumen region based on purely QHCF segmented results, we obtained the contour points of target region as shown in Fig. 8.10(d). It is obvious that some sharp corners on the top-left part of the contour and the bottom part are also not very smooth; this conflicts with normal observation of lumen shape in anatomy. In the next step, this contour was split into six equal sections and we searched the control points (see Fig. 8.10(e)) with maximum reliability criterion. The MPA algorithm was then used to track the whole contour and the result is shown in Fig. 8.10(f). Compared with the contour in Figs. 8.10(d) and 8.10(f), the effect of smoothness constraint in MPA algorithm is demonstrated. The two rough parts in Fig. 8.10(c) have also been fine-tuned.

0 0

Post a comment