Color image processing algorithms are first developed using a high-level mathematical modeling language. Current integrated development environments offer libraries of intrinsic functions, which on one hand enable faster development, but on the other hand hide the use of fundamental operations. The latter have to be detailed for an efficient hardware and/or software physical implementation. Based on the experience accumulated in the process of implementing a segmentation algorithm, this thesis outlines a design for implementation methodology comprised of a development flow and associated guidelines.
The methodology enables algorithm developers to iteratively optimize their algorithms while maintaining the level of image integrity required by their application. Furthermore, it does not require algorithm developers to change their current development process. Rather, the design for implementation methodology is best suited for optimizing a functionally correct algorithm, thus appending to an algorithm developer's design process of choice.
The application of this methodology to four segmentation algorithm steps produced measured results with 2-D correlation coefficients (CORR2) better than 0.99, peak-signal-to-noise-ratio (PSNR) better than 70 dB, and structural-similarity-index (SSIM) better than 0.98, for a majority of test cases.
Library of Congress Subject Headings
Image processing--Data processing; Image processing--Digital techniques; Computer algorithms--Evaluation
Electrical Engineering (MS)
Department, Program, or Center
Electrical Engineering (KGCOE)
Whitesell, Jamison D., "Design for Implementation of Image Processing Algorithms" (2014). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus