REAL-TIME PROCESSING OF INFRARED IMAGES: ENHANCEMENT AND TARGET DETECTION (TWO-DIMENSIONAL, FFT)

The field of digital image processing has been well established for more than ten years. During this time it has continued a rapid growth into new application areas, and techniques have improved in existing ones. Image enhancement is one fundamental application of the discipline; computer vision is an eventual goal.;An evaluation of several image enhancement algorithms was carried out to judge their effectiveness in a real-time display system. The enhancement algorithms include adaptive contrast enhancement filtering, highpass filtering, homomorphic filtering, variable threshold zonal filtering and histogram equalization. Experimental results are presented including pictures of enhanced images for a variety of infrared scenes.;Some theoretical aspects of two-dimensional discrete Fourier transforms are given. An analysis of the problem of filtering non-square images is presented, and the usefulness of an elliptically symmetric two-dimensional transfer function demonstrated. An algorithm is presented that takes advantages of the redundancies found in the transform of an image (or any real function). It reduces processing time for two-dimensional FFT's by a factor of two over the standard row-column method.;The detection approach that was explored utilized a one-dimensional view of data that is essentially three-dimensional. A sparse impulse response filter was applied to this data where most of the filter coefficients were forced to be zero. Existing one-dimensional predictor models were used to assign the non-zero weights. The models included an adaptive Widrow filter. Results of simulated real-time detection processing are presented.;The goal of this dissertation is to study the use of digital image processing techniques on a class of infrared images. Thermal infrared radiation offers an attractive system for outdoor surveillance systems for several reasons: (1) The far-infrared band contains the dominant radiation of earth temperature objects. (2) Emissivity differences produce good contrast from varying regions of most objects. (3) Good facsimiles of visual band scenes can be produced by today's imaging systems. (4) It is more difficult to camouflage temperature differences than visual band reflectance differences.