lfovia/nrvqa-nstss

No-Reference Video Quality Assessment Using Natural Spatiotemporal Scene Statistics

No-Reference Video Quality Assessment Using Natural Spatiotemporal Scene Statistics

You can find full access to our paper here.

Citation:

If you are using the code/model/data provided here in a publication, please cite our paper:

@ARTICLE{9059006,
author={S. V. {Reddy Dendi} and S. S. {Channappayya}},
journal={IEEE Transactions on Image Processing},
title={No-Reference Video Quality Assessment Using Natural Spatiotemporal Scene Statistics},
year={2020},
volume={29},
number={},
pages={5612-5624},}

This is a two-step approach i.e. spatiotemporal feature extraction followed by regression against subjective quality scores. To evaluate the performance of this approach on a given video quality assessment (VQA) dataset, we advise to follow the following steps.

Step1: Feature extraction using functions in FeatureExtraction folder.

   (i) 3D-MSCN features using SpatioTemporal_3DMSCN_Features.m
   (ii) Spatiotemporal Gabor filter based features using SpatioTemporal_I_Features.m and SpatioTemporal_Q_Features.m (both inphase and quadrature)

Step2: Performance evaluation using functions in PerformanceEvaluation folder.

   (i) VQA_using_3DMSCN.m evaluates the performance using only 3D-MSCN features.
   (ii) VQA_using_ST_features.m evaluates the performance using only spatiotemporal Gabor filter-based features.
   (ii) VQA_using_3DMSCN_and_ST_features.m evaluates the perfomance using 3D-MSCN and spatiotemporal Gabor filter based features.

Details about other folders.

   (i) videos: You can place videos here to extracts its features.
   (ii) src: Contains 3D MSCN function, AGGD parameter estimate, and Gabor 3D Kernal.
   (iii) matlab_yuvread: Contains functions to read .yuv videos.
   (iv) logistic fitting: Contain logistic regression function to find LCC and SRCC between objective and subjective scores.

nrvqa-nstss's Issues

MSU Video Quality Metrics Benchmark Invitation

Hello! We kindly invite you to participate in our video quality metrics benchmark. You can submit NRVQA-NSTSS to the benchmark, following the submission steps, described here. The dataset distortions refer to compression artifacts on professional and user-generated content. The full dataset is used to measure methods overall performance, so we do not share it to avoid overfitting. Nevertheless, we provided the open part of it (around 1,000 videos) within our paper "Video compression dataset and benchmark of learning-based video-quality metrics", accepted to NeurIPS 2022.

Would you please release the features?

Hi,

Thansk for your sharing of the code. The convolution operation in Gabor computation is time-consuming. Would you please realese the extracted features for the databases? Especially in the UGC-VQA database (KoNViD, LIVE-VQC, and YouTube-UGC).