- Get the key points from the SIFT algorithm
- From our previous experiments we know that the synapses points are more likely to occur in the darker regions of the image.
- The points derived in step one are filtered based on binary image obtained by thresholding the CLAHE image and dilating the binary image by a circular disk of know size (diameter 7 pixels). The binary image has 99% of the synapses points)
- Another optional step could be done by filtering the key points based on the scale value obtained in the step 1
- After this points lying in the edges of the image are eliminated. This is done so that region statistics could be calculated for the largest sized disk
- After obtaining the points from step 2 the following features are constructed. Disk of various sizes are considered and the following features are constructed for all the disk sizes
- Histogram of the circular region around the SIFT key point
- CDF of the same circular region
- Hu Image invariant moments.
Initial number of key descriptor points generated by the sift algorithm (Step 1) is 447,179. After running step 2
>>indices=reduceSiftPoints('/usr/sci/crcnsdata/CRCNS/Synapses/data/roiExport3/Layer1_0_0_card_resize_p25_clahe64.tif', frames);
in the /usr/sci/crcnsdata/CRCNS/Synapses/Code/Matlab/sift directory I get 87,681 key points. This filters of the SIFT key points that don't lie on the darker regions of the image. Darker regions are the places where the synapses cloud like features are visible. The scale based filtering is not done since it has not been completely studied. Then the circular regions around the key points are considered and the Hu moments are constructed for different sized disks. The entire algorithm lasted some 41 minutes. The algorithm includes the calculation of the distance between the key point and its nearest synapses point in the ground truth XML file. The algorithm can be executed with the following arguments
>>[regionAttributes nearestSynapsesDistance] = momentCalc6('/usr/sci/crcnsdata/CRCNS/Synapses/data/roiExport3/Layer1_0_0_card_resize_p25.tif', '/usr/sci/crcnsdata/CRCNS/Synapses/data/roiExport3/Layer1_0_0_card_resize_p25_clahe64.tif', '/usr/sci/crcnsdata/CRCNS/Synapses/data/roiExport3/Layer1_0_0_rois.roi_resize_p0.25.xml', frames, [5 10 15 20 25]);
this algorithm is stored in the following directory /usr/sci/crcnsdata/CRCNS/Synapses/Code/Matlab/featureGeneration
No comments:
Post a Comment