Changelog for BM3D Python / Matlab (https://webpages.tuni.fi/foi/GCF-BM3D/#ref_software)

python 4.0.2 // matlab 4.0.2

Updated BM4D dependency to version 4.2.4 for compatibility with macOS M-series ARM binaries.

python 4.0.1 // matlab 4.0.1

Fix PSD dimensionality issue with multichannel. Update BM4D dependency.

python 4.0.0 // matlab 4.0.0

Both packages now depend on "bm4d" for binaries. Removes the OpenBLAS dependency and adds much better multithreading support and the option for sharpening.
Number of threads is controlled by profile variable "num_threads". Multithreading causes slight variation in output due to numerical precision of addition, which can be eliminated by num_threads=1. Sharpening is disabled by default and can be adjusted by profile function set_sharpen.

python 3.0.9 // matlab 3.0.9

Update all references to the TIP paper, adjust block shifting and fix a binary bug related to using some simplified variances.

python 3.0.7 // matlab 3.0.7
 
Fix parameter selection bug with small image size and adjust the profile. Add an error message when trying to denoise images smaller than block size.


python 3.0.6 // matlab 3.0.5

Input/output image data is no longer restricted to range [0, 1].


python 3.0.5 // matlab 3.0.4

Wiener scaling parameter is now called 'mu2', and refiltering parameters are indicated with a '_re' naming extension.
Demo noise kernel generation logic now follows submitted manuscript.


python 3.0.4

Python binaries for MacOS now include relevant OpenBLAS functions, and should work without a separate OpenBLAS installation.


python 3.0.3 // matlab 3.0.3

Rewrote README.

Python: 
bm3d_color is now called bm3d_rgb to avoid confusion about the functionality.
Fixed some problems with custom profiles and wavelets and simplified OpenBLAS handling.


python 3.0.2 // matlab 3.0.2

Multichannel and rgb image handling is now better commented on the BM3D functions & demo files.
Python:
Simplified package contents, removed unused binaries.