BBAND Index and Adaptive-Debanding Filter

This repository contains reference matlab implementations of the BBAND Index and Adaptive-Debanding Filter from the following papers:

1. Z. Tu, J. Lin, Y. Wang, B. Adsumilli and A. C. Bovik, "BBAND INDEX: A NO-REFERENCE BANDING ARTIFACT PREDICTOR," ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Barcelona, Spain, 2020, pp. 2712-2716, doi: 10.1109/ICASSP40776.2020.9053634 IEEE Explore.
2. Z. Tu, J. Lin, Y. Wang, B. Adsumilli, and A. C. Bovik. "Adaptive Debanding Filter". IEEE Signal Processing Letters, Volume 27. IEEE Explore.

Running the code

There are third_party routines in "third_party/matlabfns". You will need to add that directory to your path (e.g., path(path(), "${PATH_TO_ROOT}/third_party/matlabfns")).

The code for the blind banding detector (BBAND) and the adaptive debanding filter are in the following files.

src/BBAD_I.m
src/deband_filter_yuv420p

Please see the comments in the files for description of the input/output parameters.

Alternatively, you can have a look a scripts used to run experiments in the papers:

1. The script for running banding detection experiments is located in src/video_band_detect.m.
2. Debanding filtering as preprocessing experiments are in src/video_preproc_dither.m and the script for running adaptive debanding postprocessing is src/video_postproc_deband_filter.m.

Tests

There are also a few tests in the "tests" folder that illustrate how to call the routines. You can run the tests from the by adding the "src" directory to your path.

test_BBAD_I
test_deband_filter

The scripts will output whether the implementation passed or failed. E.g., at the end of the debanding script you should see lines like:

KristenAndSara_10frames_512x256 passed