Abstract: A device, system and method of image processing and compression are preferably used with second generation infrared (IR) detector arrays where the pixel size is less than the optical blur of the system. Raw input pixels are compressed such that output pixels are more closely matched to the optical blur of the system. Further, the number of output pixels are compressed from a greater number of input pixels, based upon a selected input-to-output compression ratio. The ratio is not constrained to be of an integer value. The compression takes place using a plurality of neighborhood processors, each corresponding to a different output pixel type. Each neighborhood processor contains a plurality of different predetermined weighed fractions of input pixel data and are used to produce compressed output pixel data, based upon the selected input-to-output compression ratio and the output pixel type, for real-time processing.

Abstract: An improved target sighting device operable to converge on targets whose initial velocity has caused its image to move a significant fraction within the field of view during the period between detections utilizing less electrical circuitry and only one focal plane detector array. A detector array and associated signal processing system for incorporation within an electro-optical tracking system, the detector array comprising both fine track and coarse track detector elements wherein fine track and coarse track detector elements are further interconnected to hybrid fine track and coarse track signal processing algorithms. The resultant signals from the fine track and coarse track detections on the hybrid array will result in commands to gimbals about reflecting means or about a receiver telescope allowing the reflective means or receiver telescope means to be repositioned towards the target with greater accuracy.

Abstract: The selection and designation of an aimpoint on a known type of target appearing in the field of view of a video tracking system imaging sensor is accomplished by (a) developing binary target patterns from quantized sensor data generated in the video tracking system video preprocessor; (b) measuring the moments of the target patterns; (c) developing and storing a target reference pattern data base; (d) computing from real time binary coding and moment measurement data and from the stored target reference pattern data base an ensemble of estimated aimpoints; (e) testing the validity of the estimated aimpoints; and (f) averaging valid estimated aimpoints to provide an output aimpoint signal. Mechanization of the invention comprehends the use of video tracking system functions to provide quantized binary coded target patterns and the use of dedicated accumulators to generate pattern moment measurements.

Abstract: A video signal preprocessor for discriminating a target image from background in the video frame signals generated by an electro-optical sensor electrically representing the field of view thereof is disclosed. A predetermined portion of each video frame signal is selected and divided into a predetermined number of elements. Each element is assigned a gray level value from a plurality of discrete gray level values as a measure of video intensity. A discriminant function of gray level values is formed from the video data of an initial video frame signal based on the number of elements assigned to each discrete gray level value. Elements which have assigned gray level values corresponding to those of the formed discriminant function are then selected from the video frame signals subsequent to the initial one and may represent a substantial image of the target. Further preprocessing may be performed by categorizing the gray level values of the discriminant function into a set of discrete modes.

Abstract: The conversion of simple "1" and "0" PIXELS of a binary pattern into PIXELS with numerical values dependent upon relative location in the original pattern is realized by a technique that identifies various types of PIXELS including the edge PIXELS in contiguous groups of "1" PIXELS. A multilevel pattern using the various types of PIXELS is developed and a numerical value assigned to each PIXEL. The numerical value for each PIXEL is a function of its own type and the types of its near neighbor PIXELS. Pattern measurements for the augmented pattern are less sensitive to erroneous cell deletions than for simple binary patterns. The technique is implemented by a simple mechanization for generating the augmented pattern. The mechanization is characterized by operations that are easily implemented in a real time environment using standard digital logic devices.