![Electronic Thesis and Dissertation Repository](../../assets/md5images/943042f88f25dd8adfffc3aeec6e330c.png)
Thesis Format
Monograph
Degree
Master of Science
Program
Computer Science
Supervisor
El-Sakka, Mahmoud
Abstract
Image denoising is one of the most important pre-processing steps prior to wide range of applications such as image restoration, visual tracking, image segmentation, etc. Numerous studies have been conducted to improve the denoising performance. Block Matching and 3D (BM3D) filtering is the current state-of-the-art algorithm in image denoising and can provide better denoising performance than other existing methods. However, still, there is scope to improve the performance of BM3D. In this thesis, we have pointed out some of the significant aspects of this algorithm which can be improved and also suggested some approaches to get better denoising performance. We have suggested using an adaptive window size rather than the fixed window size. In addition, we have also suggested using gradient image in the blockmatching step to better facilitate the similar patch searching. Experimental results show that our suggested approaches can produce better results than BM3D irrespective of the types of image.
Summary for Lay Audience
Pixel values of natural images get corrupted by noise mostly in transmission and acquisition steps. It is needed to denoise or estimate the true pixel values from noisy pixel values for sophisticated imaging applications. Block-Matching and 3D filtering (BM3D) algorithm is one of the state-of-the-art algorithms to denoise natural images. In this thesis, we aim at improving the denoising performance of BM3D even further. BM3D uses a fixed approach (3D transformation) for the whole image. In this thesis, we have proposed an adaptive way to choose between two techniques (2D transformation and 3D transformation) for appropriate scenarios.
Recommended Citation
Zaman, Zaied, "Optimizing the usage of 2D and 3D transformations to improve the BM3D image denoising algorithm" (2019). Electronic Thesis and Dissertation Repository. 6523.
https://ir.lib.uwo.ca/etd/6523