Dear researchers,
Recently, I have been working on detecting and removing specular highlights from natural and medical images. The obtained results seem quite satisfactory. However, given that there are no clean images for my dataset, it is quite difficult to evaluate the performance of my method quantitatively. That is to say, some commonly used evaluation metrics such as PSNR, SSIM, etc. cannot be used. Please, what approach can I used to perform quantitative evaluation?
Any suggestions or comments would be highly appreciated.
With kind regards,
Patrice Monkam
If other metrics like PSNR and SSIM have been used previously in the literature, did you check how they obtained their ground truth image? Also, if you have access to any effective denoising algorithm (proprietary software), you can use that to denoise your noisy data to obtain ground truth for your proposed method. In as much as the method can outperform your proposed method. I hope this information is helpful.
Dear Patrice Monkam,
The question as to how to assess the quality of specular highlight removal results is an essential question. If you just google I believe that you can find very many litterature that deal with the subject as to how to assess the quality of spectacular higlight removal results . I recommend it to be the best method of trying to find an answer to your question from different perspectives.
Best Regards
Desalegn Abraha
Desalegn Abraha Gebrekidan Dear Prof. Desalegn Abraha Gebrekidan ,
Thank you very much for your reply! As I stated in my question, my case scenario is quite different from that of existing works given that there are no ground truths. I had tried to google, and I could not find anything related to my question.
Thank you for your time and suggestions!
Kind regards,
Patrice Monkam
Komolafe Emmanuel Dear Dr. K. Emmanuel,
I am very grateful for your suggestions and comments. I have thought of that, but it did not work in my case.
Best regards,
Patrice Monkam
Dear Dear Patrice Monkam,
Thank you very much for your comments on my recommendation.
Best Regards
Desalegn Abraha
- Badges
- Science topic
- Similar topics
- Comment