Self-Supervised Image Denoising Using Implicit Deep Denoiser Prior
Authors: Huangxing Lin, Yihong Zhuang, Xinghao Ding, Delu Zeng, Yue Huang, Xiaotong Tu, John Paisley
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate IDDP on various denoising tasks involving both synthetic and real-world noise. Synthetic Noise We collected 4744 images from the Waterloo Exploration Database (Ma et al. 2016) to synthesize noisy images for training. Several state-of-the-art denoising methods are adopted for performance comparison, including a model-based method BM3D (Dabov et al. 2007), selflearning methods DIP (Ulyanov, Vedaldi, and Lempitsky 2018), Noise2Void (N2V) (Krull, Buchholz, and Jug 2019), Self2Self (S2S) (Quan et al. 2020), Neighbor2Neighbor (Nb2Nb) (Huang et al. 2021), CVF-SID (Neshatavar et al. 2022) and Blind2Unblind (B2U) (Wang et al. 2022). BSD300 (Martin et al. 2001) is the test set of the following experiments. We adopt peak signal to noise ratio (PSNR) and structural similarity index (SSIM) as evaluation metrics. Quantitative results are reported in Table ??. IDDP achieves better denoising results than BM3D, Nb2Nb and B2U. Visual results are shown in Figure 6. |
| Researcher Affiliation | Academia | Huangxing Lin1, Yihong Zhuang1, Xinghao Ding1, Delu Zeng2, Yue Huang1, Xiaotong Tu1*, John Paisley3 1School of Informatics, Xiamen University, China 2School of Mathematics, South China University of Technology, China 3Department of Electrical Engineering, Columbia University, USA |
| Pseudocode | No | The paper does not include a dedicated pseudocode block or algorithm listing. The methodology is described using equations and textual explanations. |
| Open Source Code | No | The paper does not provide any explicit statement or link to open-source code for the described methodology. |
| Open Datasets | Yes | We collected 4744 images from the Waterloo Exploration Database (Ma et al. 2016) to synthesize noisy images for training. BSD300 (Martin et al. 2001) is the test set of the following experiments. We further evaluate the performance of IDDP on a real Fluorescence Microscopy Denoising (FMD) dataset (Zhang et al. 2019b). |
| Dataset Splits | No | The paper mentions "training set" and "test set" but does not explicitly describe a separate "validation set" or its split details. |
| Hardware Specification | No | The paper mentions training a U-Net with PyTorch and Adam but does not specify any hardware details like GPU/CPU models or memory. |
| Software Dependencies | No | The paper states, "We use Py Torch and Adam (Kingma and Ba 2014)" but does not provide specific version numbers for these or other software dependencies. |
| Experiment Setup | Yes | The training images are randomly cropped into 128 128 patches before being input to the network. The learning rate is fixed to 0.0002 for the first 1,000,000 iterations and linearly decays to 0 for the next 1,000,000 iterations. We use Py Torch and Adam (Kingma and Ba 2014) with a batch size of 1 to train the network. |