Thesis Details
Approximate Implementation of Arithmetic Operations in Image Filters
This Master's thesis deals with approximate implementations of arithmetic operations in image filters. In particular, it uses approximation techniques to adjust the multiplication operations in a non-trivial image filter. Several methods are employed, such as converting the floating-point multiplication to fixed-point multiplication, applying evolutionary algorithms, especially Cartesian genetic programming, to create new approximate multipliers that have an acceptable level of error, and at the same time, reduced filtering complexity. The result is a collection of approximate multipliers evolved with respect to the data distribution retrieved from the image filter. Approximate image filters that use evolved approximate multipliers are compared with the standard image filter on a set of images.
approximate computing, evolutionary algorithm, Cartesian genetic programming, fixed-point arithmetics, non-local denoising filter, approximate multipliers.
Beran Vítězslav, doc. Ing., Ph.D. (DCGM FIT BUT), člen
Bidlo Michal, doc. Ing., Ph.D. (DCSY FIT BUT), člen
Lengál Ondřej, Ing., Ph.D. (DITS FIT BUT), člen
Rozman Jaroslav, Ing., Ph.D. (DITS FIT BUT), člen
Zbořil František V., doc. Ing., CSc. (DITS FIT BUT), člen
@mastersthesis{FITMT23858, author = "Mat\v{e}j V\'{a}lek", type = "Master's thesis", title = "Approximate Implementation of Arithmetic Operations in Image Filters", school = "Brno University of Technology, Faculty of Information Technology", year = 2021, location = "Brno, CZ", language = "english", url = "https://www.fit.vut.cz/study/thesis/23858/" }