input:the noisy image to work on, in camera rgb scaled to [0,1] range but not clipped at 0
output:the denoised image
w:the weights for the neural network
