Poisson noise to the CT image using Pytorch Tensor

[MayThet (Sophia)]

I have applied Poisson noise to the CT image using the following code. However, I’ve found that the noise is fewer when I added the noise_sigma value of 50 ( lambda parameter in Poisson noise set to 50). I would appreciate your guidance and suggestions on another methods for incorporating Poisson noise into the neural network using PyTorch tensors.

imorig = torch.FloatTensor(imorig)
if args[‘add_noise’]:
noise = torch.poisson(torch.ones(imorig.size()) * args[‘noise_sigma’])
imnoisy = imorig + noise

[MayThet (Sophia)]

I have edited the code as follows. The following image is noise sigma 50. (Poisson noise lambda parameter = 50).

Could you please confirm if it’s correct or suggest any improvements?

if args[‘add_noise’]:
imnoisy = add_poisson_noise(imorig, args[‘noise_sigma’])

def add_poisson_noise(image_tensor, noise_factor):
noisy_tensor = torch.poisson(image_tensor *
noise_factor) / noise_factor
return noisy_tensor