Multi-View Learning with Context-Guided Receptance for Image Denoising
Authors: Binghong Chen, Tingting Chai, Wei Jiang, Yuanrong Xu, Guanglu Zhou, Xiangqian Wu
IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The model is validated on multiple real-world image denoising datasets, outperforming the state-of-the-art methods quantitatively and reducing inference time up to 40%. Qualitative results further demonstrate the ability of our model to restore fine details in various scenes. The code is publicly available at https://github.com/Seeker98/CRWKV. ... Section 4 Experiments |
| Researcher Affiliation | Academia | Binghong Chen1 , Tingting Chai2 , Wei Jiang1 , Yuanrong Xu2 , Guanglu Zhou2 and Xiangqian Wu2,3 1School of Mathematics, Harbin Institute of Technology, China 2Faculty of Computing, Harbin Institute of Technology, China 3Suzhou Research Institute, Harbin Institute of Technology, China EMAIL, EMAIL |
| Pseudocode | Yes | The detailed algorithm of CTS is presented in Algorithm 1. ... Algorithm 1 Context-guided Token Shift Input: input x, offset dictionary D, learnable weight ̖ Parameter: channel C Output: shifted output CTS(x) 1: Let psum = 0, c = 0, o = zeros(x.shape) 2: for offset p in D do 3: calculate p s Manhattan distance dp = dm(p, 0) 4: calculate offset p s weight wp = 1/dp 5: psum += wp 6: end for 7: calculate channel expansion factor k = C/psum 8: for offset p in D do 9: fill o w/ shifted x: o[c : c+k wp] = xp[c : c+k wp] 10: c = c + k wp 11: end for 12: return CTS(x) = ̖ o + (1 ̖) x |
| Open Source Code | Yes | The code is publicly available at https://github.com/Seeker98/CRWKV. |
| Open Datasets | Yes | We use the SIDD dataset [Abdelhamed et al., 2018], consisting of 320 real-world images, as the primary training set. From each high-resolution image (256 256), we crop 300 non-overlapping slices, generating a total of 96,000 training samples. For testing, we evaluate on the SIDD, ccnoise [Nam et al., 2016], and Poly U [Xu et al., 2018] datasets, all containing realistic noise. Additionally, we create a synthetic dataset, Urban100GP, by introducing mixed Additive White Gaussian Noise (AWGN) with ̖ = 10 and Poisson noise to Urban100 [Huang et al., 2015] to simulate real-world noise. |
| Dataset Splits | No | We use the SIDD dataset [Abdelhamed et al., 2018], consisting of 320 real-world images, as the primary training set. From each high-resolution image (256 256), we crop 300 non-overlapping slices, generating a total of 96,000 training samples. For testing, we evaluate on the SIDD, ccnoise [Nam et al., 2016], and Poly U [Xu et al., 2018] datasets, all containing realistic noise. The paper specifies the creation of 96,000 training samples from SIDD, and states testing on SIDD, ccnoise, and Poly U datasets. However, it does not provide explicit proportions or counts for validation/test splits for the primary SIDD dataset or how the test sets from SIDD are defined relative to the training set for reproducibility. |
| Hardware Specification | Yes | All experiments are conducted on a single NVIDIA RTX 4090 GPU, running Ubuntu 22.04 with PyTorch 2.5 as the software environment. |
| Software Dependencies | Yes | All experiments are conducted on a single NVIDIA RTX 4090 GPU, running Ubuntu 22.04 with PyTorch 2.5 as the software environment. |
| Experiment Setup | Yes | During training, images are cropped to 128 128, and data augmentation techniques such as rotations (90, 180, 270) and random flipping are applied to enhance model robustness. The training process is carried out with a batch size of 4 for a total of 288,000 iterations. We use the AdamW optimizer with ̖1 = 0.9 and ̖2 = 0.999. The learning rate starts at 3 104 and is gradually reduced to 1 106 after the 192,000-th iteration. For model-specific configurations, the output channel size of the input projection is set to 48. The depths of the four stages are empirically chosen as L1 = 3, L2 = L3 = 4, and L4 = 6. |