I will soon be posting on the 3D segmentation I was able to do, only to a certain extent, using a recursive region growing algorithm. The criterion for region growing is simple and I am using an upper and lower threshold. This is known to produce "holes" in the MIP rendering of the segmented region. It seemed quite clear later on as to why such local discontinuity occurs in the rendering. Since we are doing a Maximum Intensity Projection, the maximum intensity function around a neighborhood of the segmented region is not guaranteed to be continuous or approximately continuous.
Using a recursive region growing algorithm with a criterion like the one I am using at the moment (thresholds) is just the beginning. It is popular to use other criterions, such as the Fischer's criterion. Most of these really boil down to using some sort of statistical mesaures. I am looking to first try and see how well it works if I used a simple idea of a pixel fulfilling a group if it is within some n number of standard deviations.
Apart from research work, I have been busy preparing for finals for the Machine vision and Advanced graphics and Visualization courses I took this year. Imperial exams are very thought provoking, much like the ones I "reckon" I used to get at U. Toronto.
Using a recursive region growing algorithm with a criterion like the one I am using at the moment (thresholds) is just the beginning. It is popular to use other criterions, such as the Fischer's criterion. Most of these really boil down to using some sort of statistical mesaures. I am looking to first try and see how well it works if I used a simple idea of a pixel fulfilling a group if it is within some n number of standard deviations.
Apart from research work, I have been busy preparing for finals for the Machine vision and Advanced graphics and Visualization courses I took this year. Imperial exams are very thought provoking, much like the ones I "reckon" I used to get at U. Toronto.
No comments:
Post a Comment