Accession Number : ADA289252

Title :   ROC Analysis of IR Segmentation Techniques.

Descriptive Note : Master's thesis,

Corporate Author : AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING

Personal Author(s) : Harrup, Georgia K.

PDF Url : ADA289252

Report Date : DEC 1994

Pagination or Media Count : 112

Abstract : Receiver Operating Characteristic (ROC) curves are used to compare the effectiveness of IR image processing techniques. Two non-parametric error estimation techniques (k-Nearest Neighbor and Parzen Window) are used to create estimates of the probability density functions for the data. These pdfs are used in the creation of the ROC curves for both resubstitution and leave-one-out estimates. These estimates generate the upper and lower bounds, respectively, on the ROC curves. The ROC curve analysis is performed on the outputs of various image processing techniques and the resulting ROC curves are used to compare the techniques. Of the image processing techniques used in this thesis, the close minus open (CMO) morphological filter operation produced the best results.

Descriptors :   *IMAGE PROCESSING, *X RAY DIAGNOSTICS, *INFRARED IMAGES, *CANCER, FOURIER TRANSFORMATION, TARGET RECOGNITION, MORPHOLOGY, THESES, PROBABILITY DENSITY FUNCTIONS, PIXELS.

Subject Categories : Medicine and Medical Research
      Cybernetics

Distribution Statement : APPROVED FOR PUBLIC RELEASE