search

Vandermode / ELD

Physics-based Noise Modeling for Extreme Low-light Photography (CVPR 2020 Oral & TPAMI 2021)
http://openaccess.thecvf.com/content_CVPR_2020/html/Wei_A_Physics-Based_Noise_Formation_Model_for_Extreme_Low-Light_Raw_Denoising_CVPR_2020_paper.html
MIT License
494 stars 66 forks source link

Artifacts in the synthetic low light clean image #9

Open ProNoobLi opened 4 years ago

ProNoobLi commented 4 years ago

image Hi, I got the artifacts when generating synthetic low light clean images. According to your paper, the fake low light clean images = long exposure images / ratio, while actually such operation(large integers are divided by ratio then make float to integers) squeezes the range of the values which loses the accuracy and generates the "non-continuous step" in the image, which feels like a HDR image displayed on an 8-bit screen. The result is as follows: image original long exposure image image synthesize low light clean image after auto-brightness for imshow image original low light noisy image image synthesize noise image based on the "non-continious" low light clean image

How do you fix the artifacts?

Vandermode commented 4 years ago

while actually such operation(large integers are divided by ratio then make float to integers)

pls don't convert the float to int in this step

ProNoobLi commented 4 years ago

Hi, but anyway the photon-electrons map converted from the low-light-clean raw are integers, right??

My way: long-exposed-clean raw(integer) -> synthetic-low-light-clean(float)-> photon-electrons map(integer)-> poisson noisy photon-electrons map(integer) ->poisson noisy raw(integer)

It's not practical to keep float number in the step of generating the Poisson noisy photon-electrons map