Thesis (Selection of subject)Thesis (Selection of subject)(version: 285)
Assignment details
   Login via CAS
GPU Acceleration of Advanced Image Denoising
Thesis title in Czech: GPU akcelerace pokročilé metody odšumování obrazu
Thesis title in English: GPU Acceleration of Advanced Image Denoising
Key words: odstraňování obrazového šumu, BM3D, paralelní, GPGPU, CUDA
English key words: image denoising, BM3D, parallel, GPGPU, CUDA
Academic year of topic announcement: 2014/2015
Type of assignment: Bachelor's thesis
Thesis language: angličtina
Department: Department of Software Engineering (32-KSI)
Supervisor: RNDr. Martin Kruliš, Ph.D.
Author: hidden - assigned and confirmed by the Study Dept.
Date of registration: 03.10.2014
Date of assignment: 03.10.2014
Confirmed by Study dept. on: 18.11.2014
Date and time of defence: 07.09.2015 00:00
Date of electronic submission:31.07.2015
Date of submission of printed version:31.07.2015
Date of proceeded defence: 07.09.2015
Reviewers: Mgr. Oskár Elek, Ph.D.
 
 
 
Guidelines
Block-matching and 3D filtering (BM3D) is a well known image denoising method. It is based on locally sparse representation of an image in transform domain. This sparsity is enhanced by grouping similar image fragments into 3D array. Unfortunately, this method is both computationally and memory intensive. Despite its great denoising performance, ordinary image processing software do not employ this method, because of the high memory consumption and long computational times on common PC architectures.

Similarly to other image processing algorithms, the BM3D image denoising method presents many opportunities for data parallelism where a routine is executed over many data elements. The data parallelism can be implemented by the means of SIMD (vector) instructions, multicore CPUs, or specialized parallel accelerators such as GPGPUs and Xeon Phi cards. The main objective of this work is to design a GPU accelerated version of BM3D method and implement a prototype solution. The prototype implementation will be evaluated by a set of performance tests which will assess the applicability of GPGPUs for this particular domain and measure a relative speedup to existing CPU implementations.
References
Marc Lebrun, An Analysis and Implementation of the BM3D Image Denoising Method, Image Processing On Line, 2 (2012), pp. 175–213. http://dx.doi.org/10.5201/ipol.2012.l-bm3d

K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, "Image denoising with block-matching and 3D filtering," Proc. SPIE Electronic Imaging '06, no. 6064A-30, San Jose, California, USA, January 2006.

David B. Kirk, Wen-mei W. Hwu: Programming Massively Parallel Processors, Second Edition: A Hands-on Approach, 2012, ISBN: 0124159923

Jason Sanders, Edward Kandrot: CUDA by Example: An Introduction to General-Purpose GPU Programming, NVIDIA 2010, ISBN: 0-13-138768-5

CUDA C Programming Guide, NVIDIA Corporation, http://docs.nvidia.com/cuda/cuda-c-programming-guide/
 
Charles University | Information system of Charles University | http://www.cuni.cz/UKEN-329.html