Image denoising via a non-local patch graph total variation

Abstract

Total variation (TV) based models are very popular in image denoising but suffer from some drawbacks. For example, local TV methods often cannot preserve edges and textures well when they face excessive smoothing. Non-local TV methods constitute an alternative, but their computational cost is huge. To overcome these issues, we propose an image denoising method named non-local patch graph total variation (NPGTV). Its main originality stands for the graph total variation method, which combines the total variation with graph signal processing. Schematically, we first construct a K-nearest graph from the original image using a non-local patch-based method. Then the model is solved with the Douglas-Rachford Splitting algorithm. By doing so, the image details can be well preserved while being denoised. Experiments conducted on several standard natural images illustrate the effectiveness of our method when compared to some other state-of-the-art denoising methods like classical total variation, non-local means filter (NLM), non-local graph based transform (NLGBT), adaptive graph-based total variation (AGTV).

Three-Dimensional Ultrasound Localization Microscopy with Bipartite Graph-Based Microbubble Pairing and Kalman-Filtering-Based Tracking on a 256-Channel Verasonics Ultrasound System with a 32 × 32 Matrix Array

“…In addition, the SNR will be slightly increased with longer pulse cycle with the cost of the degradation of the axial resolution. Furthermore, noise reduction approaches, such as non-local mean denoising [51] or debiased noise-suppression method [52] can be applied to suppress the noise effect at the expense of computational cost. Furthermore, the field-of-view and the transmitted plane-wave angle were limited due to the grating lobe of this matrix probe.…”

Joint Spine Segmentation and Noise Removal From Ultrasound Volume Projection Images With Selective Feature Sharing

“…This implies that DIP [57] deteriorates the original image content. Although the TV-based algorithms, i.e., GSD [27] and NPGTV [30], achieve comparable performance with ours on SNRR, their MOS are even lower than that of the original image. This shows that the TV-based algorithms can remove noise but fail to preserve the content of the original image.…”

“…To validate the effectiveness of our proposed method on scan noise removal, we performed user studies and scan noise removal rate based on the comparison with other weakly supervised denoisers that do not require the paired cleannoisy images for learning, such as Noise2Void (N2V) [37], Deep Image Prior (DIP) [57], Cycle-GAN (CYC) [58], and the total variance based denoising algorithms such as Graphbased Sinogram Denoising (GSD) [27], and non-local patch graph total variation (NPGTV) [30], as well as our previous work DAGAN [16]. In the user study, we randomly selected 15 testing images, and invited 8 ultrasound experts to assess the restored image quality.…”

“…Similar to ultrasound scan noise, Computerized Tomography (CT) and Electron Tomography (ET) reconstructions usually encounter significant amounts of noise caused by low dose, limited, and fragmented data. Total variation (TV) regularization, which minimizes the total variance of an image while preserving the content of the original image, has been used extensively in the noise removal of CT and ET reconstruction [27]- [30]. Mahmood [27] adopted a total variance regularization strategy directly in sinogram for tomographic reconstruction.…”

“…6: An overview of the proposed framework, which consists of four main components, i.e, a preprocessing module, a noise removal module, a spine segmentation module, and selective feature-sharing blocks. variation (NPGTV) [30] introduces the pixel coordinate as an ingredient to construct the K-nearest neighbor graph for effective denoising. We will compare the denoising results of various kinds of algorithms with those of our proposed algorithms in the experiment section.…”