krshrimali / no-reference-image-quality-assessment-using-brisque-model

Implementation of the paper "No Reference Image Quality Assessment in the Spatial Domain" by A Mittal et al. in OpenCV (using both C++ and Python)

Note

Please see #4 if you are looking to contribute. :)

Note: There have been many revisions, specially about the models used. Please check the work by @clunietp (https://github.com/clunietp) and myself on BRISQUE IQA committed to opencv_contrib here: https://github.com/opencv/opencv_contrib/tree/master/modules/quality

Note: This work was originally published on LearnOpenCV. For code: https://github.com/spmallick/learnopencv/tree/master/ImageMetrics

The work was published on https://www.learnopencv.com/. Link: blog post.

Process

Installation Instructions

Python 2.x LIBSVM Installation sudo apt-get install python-libsvm

Python 3.x LIVSVM Installation and C++ LIBSVM Installation

For C++ :

1. cd C++/libsvm/
2. cmake .
3. make

For Python 3.x :

1. cd Python/libsvm/
2. make
3. cd python
4. make

Usage

Python 2.x

1. python2 brisquequality.py <image_path>

Python 3.x

1. cd Python/libsvm/python/
2. python3 brisquequality.py <image_path>

C++

1. cd C++/
2. ./brisquequality <image_path>

Example

Quality Score Comparison using BRISUQE Model

no-reference-image-quality-assessment-using-brisque-model's Issues

This PR needs attention: Refactor code, better documentation and build script

I appreciate the attention given to this project, and I thank all of you for that. Keeping that in mind, I've decided to give this work another look. A few notes:

• This code was ported to OpenCV's contrib repository (quality module: https://docs.opencv.org/4.5.2/d8/d99/classcv_1_1quality_1_1QualityBRISQUE.html).
• This repo uses incorrect models which might result in wrong results, this was however corrected while porting to OpenCV.
• I wrote this code when I was in my 2nd year of Bachelors's, pretty young! ;) I've loved C++ more with time, and I plan to write better code for this.

So, what's the plan?

• Use models appropriate for this code. Create a link in case the models are deleted from the source (will have to check if it's allowed or not though).
• Refactor code for both C++ and Python.
• Re-visit dependency of LIBSVM for this project.
• Build scripts for the C++ code. (Cross-platform)
• Better documentation on how to use, and known issues (if any).
• Address all the opened issues on this project.

This project has come back to life, thanks to all of you!

How to contribute?

In case you are planning to contribute, please reach out here and let me know what's your plan (if you have one already), and what task you want to pick up. I'll create separate issues for the tasks above later, but for now, please feel free to comment here.

Thanks!
~ Kushashwa (/krshrimali/Kush)

Great Work! Well done!!!

Training own brisque model in python

can anyone please tell me how to train a brisque model in python?
How to generate the allmodel file?

issues

hello, I have run your code on windows system, but I meet some compilation issues.

the problems are as follows:

1.CMake Error: The source directory "C:/Users/pan/Desktop/No-Reference-Image-Quality-Assessment-using-BRISQUE-Model-master/C++/libsvm" does not appear to contain CMakeLists.txt.

2.make: Nothing to be done for 'all'.

hope that you can answer my confuses. thank you very much!

the brisque model seems yield wrong result quite often

the brisque model seems yield wrong result quite often. I have original png image, jpeg compressed with 90% quality, jpeg with 60% quality.
The computed score is:

• (23.201696395874023, 0.0, 0.0, 0.0) for png
• (19.102746963500977, 0.0, 0.0, 0.0) for high quality jpg
• (17.447845458984375, 0.0, 0.0, 0.0) for low quality jpg

If I get the paper correct, the higher the score the lower the quality.

png:

high quality jpeg:

low quality jpeg:

About the all model

First, thank you for providing the implementation of BRISQUE.
And I want to know where the "allmodel" file used by svm comes from?
Does the "allmodel" file trained from the tid2013 or tid2008 dataset?