garymataev/deepred

DeepRED: Deep Image Prior Powered by RED

DeepRED: Deep Image Prior Powered by RED

ICCV 2019, Learning for Computational Imaging (LCI) Workshop: http://openaccess.thecvf.com/content_ICCVW_2019/html/LCI/Mataev_DeepRED_Deep_Image_Prior_Powered_by_RED_ICCVW_2019_paper.html

Archive: https://arxiv.org/abs/1903.10176

You can reproduce the results in the article using this code

Abstract:

Inverse problems in imaging are extensively studied, with a variety of strategies, tools, and theory that have been accumulated over the years. Recently, this field has been immensely influenced by the emergence of deep-learning techniques. One such contribution, which is the focus of this paper, is the Deep Image Prior (DIP) work by Ulyanov, Vedaldi, and Lempitsky (2018). DIP offers a new approach towards the regularization of inverse problems, obtained by forcing the recovered image to be synthesized from a given deep architecture. While DIP has been shown to be quite an effective unsupervised approach, its results still fall short when compared to state-of-the-art alternatives. In this work, we aim to boost DIP by adding an explicit prior, which enriches the overall regularization effect in order to lead to better-recovered images. More specifically, we propose to bring-in the concept of Regularization by Denoising (RED), which leverages existing denoisers for regularizing inverse problems. Our work shows how the two (DIP and RED) can be merged into a highly effective unsupervised recovery process while avoiding the need to differentiate the chosen denoiser, and leading to very effective results, demonstrated for several tested problems.

Python 3

Pytorch >= 0.4

Follow the comments and instructions in the jupyter notebook

deepred's Issues

Could you upload your test images

Dear Gary Mataev

I downloaded the code you made public on github, and when I ran the program, I found that some of the test images used were directly using fixed paths.

For example:
in denoising.ipynb:
data_dict = load_image('datasets/CBM3D/house.png', sigma=SIGMA, plot=True)
in deblurring.ipynb:
data_dict = load_imgs_deblurring('datasets/Color NCSR/Butterfly.tif', BLUR_TYPE, NOISE_SIGMA, plot=True)

Those appear to be customized test samples.
Could you please upload your images for testing.

It would be an honor to have your help!
Best regards!

The data in the experiment

Hello, can you provide the data in the experiment, I did not find the experimental picture in the paper.

Will the image Inpainting code also be uploaded?