Hardware Architectures for the Correspondence Problem in Image Analysis

Thomas Eide Paulsen

    Research output: Book/ReportBookResearchpeer-review

    Abstract

    This dissertation describes investigations that have been made in order to find an optimized hardware architecture with respect to three typical matching processes in image analysis. The three matching processes are denoted the Token Constellation Method, the Triangle Method and the Simulated Annealing Method. They all have as their objectives to match objects in two images of the same scene, but taken from slightly different positions possibly under different lighting conditions and/or with slightly different orientations of the camera. The matching method denoted the "Simulated Annealing Method"has been developed in conjunction with the work on this thesis and has not previously been described. Also, during this project a combined image acquisition and compression board has been developed for a NASA sounding rocket. This circuit, a so-called Lightning Imager, is also described. Finally an optimized hardware architecture has been proposed in relation to the three matching methods mentioned above. Because of the cost required to physically implement and test the developed architecture, it has been decided todocument the performance of the architecture through theoretical proofs only.
    Original languageEnglish
    Place of PublicationLyngby
    PublisherIAU / DTU
    Number of pages100
    Publication statusPublished - 1996

    Cite this