Comments (8)
Thanks. The deblurred Y channel is used to replace Y component of blurry image (YCbCr space), which is then converted back to RGB space. You can try it.
I will update it along with a new way to directly deblur RGB images.
from selfdeblur.
I didn't find where do you transform the image from RGB space to YCbCr space. In the calculation of loss, out_y
([1,1,128,128]) and y
([1,3,128,128]) are directly input into the MSE loss .
total_loss = mse(out_y, y) + tv_loss(out_x)
From the source code of mse, out_y
will be broadcasted to be [1,3,128,128].
expanded_input, expanded_target = torch.broadcast_tensors(input, target)
ret = torch._C._nn.mse_loss(expanded_input, expanded_target, _Reduction.get_enum(reduction))
Therefore, my question is, since the network is not trained with the supervision of Y component of blurry image, how can we expect out_y
to be the Y component of deblurred image? @ @ @csdwren
from selfdeblur.
I have the same question.
from selfdeblur.
https://github.com/csdwren/SelfDeblur/blob/master/selfdeblur_ycbcr.py has been updated to handle color images. Also the code has been improved with better robustness. Thanks.
from selfdeblur.
In your paper, you used the non-blind deconvolution method [14], i.e. "D. Krishnan and R. Fergus. Fast image deconvolution using hyper-laplacian priors. InNIPS, 2009.". I found that the input image of this non-blind deconvolution method I found on the web must be a grayscale image, and the output image must be a grayscale image, but my final deblurring result is to get the restored color image. This problem has been bothering me for a long time. I am wondering if you could kindly send me the source program and necessary information of this part. I promise they will be used only for research purposed.
Thank you very much for your kind consideration and I am looking forward to your early reply.
from selfdeblur.
Thank you for your work. I mainly want to know how to turn the deconvolution result using "D. Krishnan and R. Fergus. Fast image deconvolution using hyper-laplacian priors. InNIPS, 2009." into a color clear image. Thank you very much !!!
from selfdeblur.
You can refer to the code 'selfdeblur_ycbcr.py'
An RGB image is converted to YCbCr, and the deconvolution result only process Y channel. Finally re-converted YCbCr to RGB.
from selfdeblur.
from selfdeblur.
Related Issues (18)
- the link of download is unvaliable HOT 1
- RuntimeError: CUDA out of memory HOT 2
- RuntimeError: set_storage_offset is not allowed on a Tensor created from .data or .detach(). HOT 2
- Can't find the training set HOT 2
- test HOT 6
- Why is PSNR lower than in the paper ? HOT 1
- About the estimating of blur kernel HOT 6
- The noise level HOT 2
- TV loss
- Colour image deblurring
- Online/offline image deblur solution HOT 1
- Pre trained models! HOT 2
- Can you please fix the onedrive link to the trained models? HOT 1
- Possible mismatch for examples from Lai's dataset HOT 1
- where is the pretrained model HOT 1
- cublas runtime error : the GPU program failed to execute at /opt/conda/conda-bld/pytorch_1533672544752/work/aten/src/THC/THCBlas.cu:411 HOT 1
- other fuzzy kernels HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from selfdeblur.