MULTI-RESOLUTION IMAGE FUSION

TWO SIGNIFICANT TRENDS IN IMAGE PROCESSING TECHNOLOGY OFFER REVOLUTIONARY PERFORMANCE IMPROVEMENTS FOR A VARIETY OF AIR FORCE APPLICATIONS. ONE IS THE DEVELOPMENT OF DATA FUSION ALGORITHMS WHICH OPERATE DIRECTLY ON VIDEO DATA FROM A SENSOR, AS OPPOSED TO CONVENTIONAL TECHNIQUES WHICH DETECT POINT TAR GETS AND PROCESS ESTIMATES OF THEIR KINEMATIC PARAMETERS. THE SECOND IS THE INTRODUCTION OF A WHOLE NEW CLASS OF OBSERVED PHENOMENA INTO WAVELET TRANSFORMS, WHICH PERMIT DECOMPOSITION OF OBSERVED PHENOMENA INTO DIFFERENT PHYSICAL SCALES. THE COMPUTATIONAL LEVERAGE PROVIDED BY MULTI-SCALE IMAGE PROCESSING TECHNIQUES, WHEN APPLIED TO MULTI-SENSOR IMAGE DATA, PROMISES TO INCREASE THE SENSITIVITY AND RELIABILITY OF IMAGE PROCESSING SYSTEMS IN SEVERAL APPLICATIONS. THIS PROPOSAL OFFERS TO EVALUATE THE FEASIBI LITY OF MARRYING MULTISCALE IMAGE PROCESSING TECHNIQUES TO MULTISENSOR IMAGE DATA. THE PRODUCT WOULD BE A SOFTWARE TESTBED WHICH GENERATES SYNTHETIC "GROUND TRUTH" IMAGE DATA, EXTRACTS SENSED DATA FOR SENSORS OPERATING AT DIFFERENT RESOLUTIONS (E.G. CORRESPONDING TO SIDELOOKING RADAR AND OVERHEAD INFRARED SENSORS), IDENTIFIES MODEL PARAMETERS NECESSARY TO PROCESS THE DATA, AND PERMITS ASSESSMENTS OF ALGORITHM PERFORMANCE. OF NECESSITY, THE MODELS USED IN THE TESTBED WILL BE NOTIONAL, CAPTURING KEY ASPECTS OF MULTIRESOLUTION SENSING SYSTEMS BUT OMITTING CALIBRATION TO SPECIFIC CASES.