Abstract

A circuit and method for computations on large groups of input operands is disclosed. More specifically, the present invention is related to focal plane processing. The method of the present invention performs parallel data computation exemplified by a low latency global characterization of the input operands and the sorting of the input operands. The circuit of the present invention is comprised of a cellular matrix which is comprised of a plurality of cells upon which the input operands are introduced. All the cells are simultaneously placed into the active state by a reset means. The duration of the active state in each of the cell is determined as a monotonic function of the magnitude of the input operand received by the cell after which the cell enters the idle state. Each cell generates an event for signaling to the outside of the cell at the instant the active state ends and the idle state begins. Any two cells receiving input radiation of different magnitudes generate events at different times. A global processing means receives the events from the cellular matrix and continuously produces a count indicating the number of the cells that are simultaneously in the active state, or alternatively in the idle state. Such a count is a temporal representation providing a detailed description a cumulative histogram of the input operands. With a memory means associated with each of the cells, this count can be sampled and stored at the instant when the associated cell generates the event. Such a stored memory content is an evolving histogram--equalized representation of the input operands and is result of sorting input operands.

Other References

• Burgi et al., "Asynchrony in image analysis: using the luminance-to-response-latency relationship to improve segmentation," J. Opt. Soc. Am. A, vol. 11, No. 6, Jun. 1994, pp. 1720-1726
• Cormen et al., "Section 9.2.: Counting Sort", Introduction to Algorithms, Sixth Printing, 1992, The MIT Press
• Forchheimer et al., "Near-Sensor Image Processing: A New Paradigm", IEEE Trans. On Image Processing, vol. 3, No. 6, Nov. 1994, pp. 736-746
• Astrom et al., "An Integrated Sensor/Processor Architecture Based on Near-Sensor Image Processing", Proceedings of the Workshop on Computer Architecture for Machine Perception, New Orleans, Dec. 15, 1993, pp. 147-154
• Astrom, "Smart Image Sensors", Ph.D. Thesis No. 319, Linkoping University, Sweden, Nov. 1993
• Vladimir Brajovic et al, "Computational Sensors for Global Operations", IUW (ARPA), Nov. 13-16, 1994, pp. 621-630
• Kanade et al. "Computational Sensors", DARPA Workshop U. Penn, May 11-12, 1993, Also in IUW '93
• Ballard & Brown: "Computer Vision", Prentice-Hall, Inc. 1982 Sect. 3.2.2. Histogram Transformations
• Ballard & Brown: "Computer Vision", Prentice-Hall, Inc. 1982 Sect. 5.3 Global Techniques: Region Growing via Thresholding
• Blelloch et al. "Comparison of Sorting Algorithms for Connection Machine CM-2", SPAA 1991
• Ishikawa et al. "High Speed Vision System Using Massively Parallel Processing", Proc. IEEE Intl. Conf. Robotics, '92
• Yamada et al. "High Speed Target Tracking Using Massively Parallely Processing Vision", Proc of IEEE Intl. Conf on Rob. '93, pp. 267-272
• Weckler: "Operatoin of P-N Junction Photodetectors in a Photon Flux Integrating Mode", IEEE, vol. SC-2, No. 3, Sep. '6