Supplementary Material PUCA: Patch-Unshuffle and Channel Attention for Enhanced Self-Supervised Image Denoising

We use the SIDD-Medium and DND datasets for training. Our minibatch consists of two augmented samples. The stride factor of test time is 2. For the post-processing strategy, we adopt the random-replacing We use the sum of the channels of the output's central pixel as the score and We visualize the denoising results for additional samples. When awgn-based denoiser meets real noises.