search

fatheral / matlab_imresize

Python implementation of MatLab imresize function
MIT License
154 stars 41 forks source link

Question about the article #1

Open chaos5958 opened 6 years ago

chaos5958 commented 6 years ago

I read your article posted at https://www.reddit.com/r/MachineLearning/comments/6vdo51/p_matlab_bicubic_imresize_implemented_in_python/.

I totally agree with you that using matlab enforces off-line dataset building. (bad!!!) You commented that "Note: it is not very optimized in terms of speed, and probably the best solution is to re-compile OpenCV (there single constant for cv::resize should be changed). But it is not always appropriate for all.", I have two questions with this one.

  1. What single constant should I change to use OpenCV bicubic similar to matlab version?
  2. Does OpenCV bicubic show similar performance (e.g. PSNR) to matlab bicubic?

Thank you, Hyunho.

fatheral commented 6 years ago

Hi Hyunho!

Concerning your first question, I gathered all information about how to tweek OpenCV here - https://github.com/fatheral/matlab_imresize#addinfo . This contains specific file, function and constant name.

As for the similar PSNR in comparison with matlab imresize, in fact, I haven't test it - had no permissions on the available server at that time. But if you could try and compare - it would be great! Please let me know if it is working.

chaos5958 commented 6 years ago

I tried opencv and scikit-image, but still different from matlab.

opencv-matlab: Matlab has additional anti-aliasing filter, and I'm not sure opencv can reproduce the same result.

scikit-image: Scikit image and matlab show different result even when I turn-off anti-aliasing filter for both ones.

I hope there is some reference interpolation written in all language (e.g. python) for super resolution research.