CUDA Accelerated Multi-domain Volumetric Image Segmentation and Using a Higher Order Level Set Method

Ojaswa Sharma, François Anton, Qin Zhang, Chandrajit Bajaj

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

    Abstract

    In this paper we present a methodology for smooth surface segmentation (partition) of volumetric images using higher order level set scheme. The segmentation allows for a multi-domain partitioning by minimizing modi-fied Mumford-Shah functional. Since, volumetric images tend to be de-manding in terms of computation and memory space, we employ a CUDA based fast GPU segmentation and provide accuracy measures compared with an equivalent CPU implementation. Our resulting surfaces are C2-smooth resulting from tri-cubic spline interpolation algorithm. We also provide error bounds on the reconstruction/segmentation.
    Original languageEnglish
    Title of host publicationISPRS International Workshop on Multidimensional & Mobile Data Model
    Publication date2009
    Publication statusPublished - 2009
    EventISPRS International Workshop on Multidimensional & Mobile Data Model - Malaysia
    Duration: 1 Jan 2009 → …

    Conference

    ConferenceISPRS International Workshop on Multidimensional & Mobile Data Model
    CityMalaysia
    Period01/01/2009 → …

    Keywords

    • GPU computation
    • smooth surface reconstruction
    • CUDA
    • multi-domain volumetric segmentation
    • Level set methods

    Fingerprint Dive into the research topics of 'CUDA Accelerated Multi-domain Volumetric Image Segmentation and Using a Higher Order Level Set Method'. Together they form a unique fingerprint.

    Cite this