search

KuldeepKulkarni / ReconNet

ReconNet: CVPR 2016
https://kuldeepkulkarni.github.io/reconnet.htm
MIT License
44 stars 25 forks source link

reproduction of results with BM3D denoiser #1

Open anilrgukt opened 7 years ago

anilrgukt commented 7 years ago

Dear Kuldeep,

I was trying to reproduce the results described in your paper with BM3D denoiser. With out BM3D I am able to get same PSNR as mentioned in the paper but with BM3D(here) I am not able to achieve the gains that you have mentioned. For sigma estimation, I followed the step mentioned in your paper in Experiments: Baselines section. Is there any scaling factor in addition to that? As BM3D code needs sigma parameter on the scale of 0 to 255.

thanks, Anil

TiantianZhang commented 7 years ago

The same problem. When I tried to dnoise with BM3D, the results perform not normal, some pixels almost disappear. Have you solved this problem?

FutureIsM commented 7 years ago

hello，i see you said you were trying to reproduce the results , and i also try to do that , but i have some puzzle to make "destrministic initialization" in the paper. Do you finish reproduce that?

TiantianZhang commented 7 years ago

I have did some works on it. But when I tried to change some parameters, something becoming wrong. And the reproduce is done.

KuldeepKulkarni commented 6 years ago

Hi Tiantian,

Sorry for the delay in responding. Were you able to reproduce results with the BM3D denoiser?