Feature detection for spatial templates

The Color Medical Image System (CMIS), a program that uses segmented mapping techniques to obtain high resolution digital images, is currently trying to develop techniques to transfer microscopic glass slides to electronic image libraries. One technique that has been attempted is to use correlation techniques to scan the image. However, when segments of high magnification are used, it is difficult and time consuming to perform correlation techniques. This project investigates feature detection in microscopic images. Various techniques are implemented to detect the section of the image containing the most feature information, thereby making the correlation process more efficient. Three tests are implemented that eliminate the background in the image and calculate the mean (1st order technique), variance (2nd order technique), and ratio test (1st order technique) of the remaining pixel values. Background elimination involves deleting all pixel values above a certain experimental value from any calculations made. The source code for each of the three tests was implemented and tested on a number of images using the green color band. Each program outputs the box containing the most features and writes that section to a file to be displayed to the screen. A visual rank was also recorded so as to compare it the output of the tests. Each of the three tests proved to be successful. After comparing the visual rank to the output of the tests, it was determined that both first and second order techniques are effective in detecting features in microscopic images. Although all of the purposes and goals were met, this investigation should be expanded to include texturized images and the use of all three color bands