search

yuanjunchai / IKC

Implementation of 'Blind Super-Resolution With Iterative Kernel Correction' (CVPR2019)
Apache License 2.0
218 stars 44 forks source link

about random_batch_noise? #4

Closed idealboy closed 5 years ago

idealboy commented 5 years ago

Sir,

Have you haven the random_batch_noise function committed? thank you!

yuanjunchai commented 5 years ago

Yeah. It has fixed. Any other problems?

idealboy commented 5 years ago

when calling SRMDPreprocessing, I found that only 'sig=2.6' was used to generate the LR images.

Should it use uniform sample for sig for each image according the original paper?

thank you very much.

idealboy commented 5 years ago

The kernel size is fixed to 21×21. When applying on real world images, we use the additive Gaussian noise with covariance σ = 15

Sorry, I think I could not make sure which parameter will represent this σ = 15, thank you ,sir!

yuanjunchai commented 5 years ago

when calling SRMDPreprocessing, I found that only 'sig=2.6' was used to generate the LR images.

Should it use uniform sample for sig for each image according the original paper?

thank you very much.

'sigma=2.6' refers to fixed test kernel, you could change the setting according to the paper.

yuanjunchai commented 5 years ago

The kernel size is fixed to 21×21. When applying on real world images, we use the additive Gaussian noise with covariance σ = 15

Sorry, I think I could not make sure which parameter will represent this σ = 15, thank you ,sir!

self.noise_high is sigma of noise

idealboy commented 5 years ago

The kernel size is fixed to 21×21. When applying on real world images, we use the additive Gaussian noise with covariance σ = 15 Sorry, I think I could not make sure which parameter will represent this σ = 15, thank you ,sir!

self.noise_high is sigma of noise

thank you , sir

nanmehta commented 2 years ago

Could u please help me what exactly is ground truth kernel? as for Super-resolution we consider ground truth image to be the actually HR image and LR image to be the downsampled version? So my doubt is what have we actually consider to be a ground truth kernel?