The images directory contains an image from an automatic karyotyping system, karyo.aiff. This contains several chromosomes on a white background. In an automatic system the first problem is to locate these chromosomes accurately prior to classification. Today we will attempt to do this via the process of binary thresholding and skeletonisation.
Load the karyo.aiff image into the image calculator. By examination of the range of grey levels within the background and chromosome data determine a suitable threshold for binarisation of the image. This can be achieved by casting the image to a binary with an appropriate choice of threshold (all values above the threshold will be set to 1 and the remainder to 0).
Now apply the binary skeletonisation skel and examine the results carefully. The skel button applies the standard algorithm as described in Gonzales and Wintz. You will probably find that although the image can be well segmented and binarised in places, in others the extracted chromosomes become fragmented. Use the erode and dilate functions to form open and closing operations (as described in the lectures). The implementation of these operations in TINA is as circular grey level morphological operators and you can apply these to either the original or binary image. The output is always a grey level image but an equivalent binary operation can be obtained by appropriate binary thresholding of the output. Can you make the final skeletonised result more reliable (i.e. eliminate gaps) using these processes?
The second set of data for this practical is the image of a retina. Here we are interested in extracting the blood vessels in the image as completely as possible. Repeat the procedures you have developed for the karyotype images with the retina.aiff image.
Finally, apply the grey level morphological operators to the familiar house image at a range of scales. You should find that by performing appropriate sequences (as described in your lectures) it is possible to eliminate specific scales of feature.