PDE Software for Digital Image Management

Period of Performance: 01/19/2000 - 03/31/2001


Phase 1 SBIR

Recipient Firm

Mathsoft, Inc.
1700 Westlake Avenue N., Suite 500
Seattle, WA 98109
Principal Investigator

Research Topics


The DoD collects large amounts of digital imagery from diverse sources such as hospital radiology departments, satellites, and unmanned surveillance vehicles. To systematically analyze these data, one must be able to denoise, segment, annotate, store, transmit, index, screen, and retrieve images in a consistent and easy-to-use environment. All of these tasks rely on image processing techniques to extract information from the raw imagery. Traditional image enhancement and segmentation techniques take a global view of the image, processing the entire image without considering local variations. Multiscale image processing techniques based on curvature-driven flows overcome many of the problems of traditional processing methods, especially for image enhancement and segmentation, since these techniques focus on features of the data that are found at different scales and resolutions. We propose to develop image enhancement, segmentation, motion estimation, and object tracking methods using curvature-based flow-driven nonlinear partial differential equation imaging models and algorithms, including total variation, level-set, and active contour techniques. We will demonstrate the feasibility of these techniques on specific applications in medical and tactical imaging, or to Army sponsor-directed imaging applications. This technology will improve the ability to analyze, index, and search digital imagery. BENEFITS: This research will lead to a software system for the processing, analysis and interpretation of digital image sequences using curvature-based flow-driven techniques including total variation, level-set, and PDE-based image techniques. The system will be demonstrated on medical imagery and on Army-sponsor supplied (tactical and/or medical) data. For DoD applications, this system will be valuable for medical and tactical image database management, medical clinical studies, target acquisition, and battlefield damage assessment. Commercial applications include medical imaging, and image and video database indexing, archiving, and retrieval.