Real Data Analysis Circular Vs Elliptical 961 Circular Binarization

Select class is used for binarization of the classified image. The frequency of each pixel value in the ROI is determined. The core class C0 is the class with the greatest number of pixels. The number of pixels is equivalent to the area of the core. The average area of the entire lumen core is determined from the ground truth boundaries, and this area is compared to the area of the C0 area. A threshold function is used to determine whether to binarize the C0 region, or to merge C0 with C1 and then binarize. We now discuss the methods to compute the average lumen area, lumen core area, and the difference of these, and their comparison. Lumen Area Computation by Triangle/Scan-Line Methods (A)

To determine the average area of the entire lumen from the ground truth boundaries, the area by triangles computation is used. The center point of the ROI is the user input, and is equivalent to the center of gravity (CG). The area of the enclosed region is obtained by summing the areas of the triangles formed by the CG and each pair of neighboring points on the boundary.

In the scan-line method, we count the number of pixels along the scan line which lies in the ROI. This process is done for all the lies which interest the ROI region. The entry and exit points are computed by finding the number of times the scan line interests the boundary yielding the odd or even number. If the intersection yields 1 then begin counting the pixels, and if the intersection yields 2, then stop counting pixels. This gives a total number of pixels along the line. The process stops when there are no more interesections. In a 384 x 512 image, the average area for the left and right lumen is 500 pixels squared.

The select class package takes as one of its inputs the number of classes formed after the segmentation method. Using this as a size for an array of the different classes C0 through Cn, the program checks each pixel in the ROI and stores the number of times that each of the different pixel values occur. The program then sorts these class values by their frequency. Difference Computation (A — B) and Comparison with Threshold

Using the average ground truth contour area, a difference threshold, Td, is determined. We set Td = 75. If the difference between the average ground truth contour area and the number of pixels of C0 in the ROI is less than the difference threshold, then only C0 is selected. If the difference is greater than the difference threshold, then both C0 and C1 are selected, then they are merged and a binary image is made.

In the GSM, a select class package is not used, but a region growing method is used. The GSM usually merges the C0 and C1 classes, so the region growing captures both C0 and C1 classes.

Comparison of 3 Classification Techniques Error Metric: Polyline Distance-Based Resolution: 1 pixel = 0.25 mm







Figure 9.44: Results using FCM, MRF, and GSM methods.

0 0

Post a comment