Overview
BM3D (Block-Matching and 3D filtering) is an advanced image and video denoising algorithm. It operates by grouping similar 2D image fragments (blocks) into 3D data arrays. Collaborative filtering is then applied, involving a 3D transformation, shrinkage of the transform spectrum, and inverse 3D transformation. This process attenuates noise while preserving fine details. Overlapping filtered blocks are aggregated using a weighted averaging procedure, leveraging redundancy for improved estimation. The algorithm offers state-of-the-art performance in terms of peak signal-to-noise ratio (PSNR) and subjective visual quality. Extensions such as BM4D and VBM3D cater to different data types like video and Rician-distributed data. The software is available for non-profit education and scientific research.
Common tasks