In traditional image denoising, noise level is an important scalar parameter which decides how much the input noisy image should be smoothed. Existing noise estimation methods often assume that the noise level is constant at every pixel. However, real-world noise is signal dependent, or the noise level is not constant over the whole image. In this paper, we attempt to estimate the precise and pixelwise noise level instead of a simple global scalar. To the best of our knowledge, this is the first work on the problem. Particularly, we propose a deep convolutional neural network named “deep residual noise estimator” (DRNE) for pixelwise noise-level estimation. We carefully design the architecture of the DRNE, which consists of a stack of customized residual blocks without any pooling or interpolation operation. The proposed DRNE formulates the process of noise estimation as pixel-to-pixel prediction. The experimental results show that the DRNE can achieve better performance on nonhomogeneous noise estimation than state-of-the-art methods. In addition, the DRNE can bring denoising performance gains in removing signal-dependent Gaussian noise when working with recent deep learning denoising methods.

Person re-identification (Re-ID) has become a hot topic in both research and industry. We joined in a person Re-ID challenge of the First National Artificial Intelligence Challenge (China, 2019) and found some model designs and training tricks work great or not on a super big private dataset. In this paper, we propose a model that combines the most effective designs, including multi-scale, multi-branch and attention mechanism, and report training tricks that are no less or even more important in improving person Re-ID performance. We analyze four commonly used public datasets: Market1501, DukeMTMC-ReID, CUHK03, and MSMT17, and achieve the state-of-the-art performance. Besides, we analyze and confirm the effectiveness of the designs by ablation studies. We also share strategies that play a key role in the challenge and experience of model designs that do not generalize well on large datasets.

With the proliferation of large-scale 3D point cloud datasets, the high cost of per-point annotation has spurred the development of weakly supervised semantic segmentation methods. Current popular research mainly focuses on single-scale classification, which fails to address the significant feature scale differences between background and objects in large scenes. Therefore, we propose MCCR (Multi-scale Classification and Contrastive Regularization), an end-to-end semantic segmentation framework for large-scale 3D scenes under weak supervision. MCCR first aggregates features and applies random downsampling to the input data. Then, it captures the local features of a random point based on multi-layer features and the input coordinates. These features are then fed into the network to obtain the initial and final prediction results, and MCCR iteratively trains the model using strategies such as contrastive learning. Notably, MCCR combines multi-scale classification with contrastive regularization to fully exploit multi-scale features and weakly labeled information. We investigate both point-level and local contrastive regularization to leverage point cloud augmentor and local semantic information and introduce a Decoupling Layer to guide the loss optimization in different spaces. Results on three popular large-scale datasets, S3DIS, SemanticKITTI and SensatUrban, demonstrate that our model achieves state-of-the-art (SOTA) performance on large-scale outdoor datasets with only 0.1% labeled points for supervision, while maintaining strong performance on indoor datasets.

The goal of image denoising is to recover a clean image from noisy input(s). For single image denoising, utilising similarities (or priors) within and across an image dataset helps recover clean images. As the noise level increases, using multiple frames become feasible, which is defined as burst denoising. In this study, the authors propose a deep residual model with squeeze‐and‐excitation (SE) modules for the burst denoising. Unlike previous methods, the authors' model does not need an explicit aligning procedure, which is light‐weighted and fast. The network contains a noise estimation convolutional neural network, which makes it capable of blind denoising. Besides, by inverting the image processing pipeline and simulating real noise in bursts, their model can suppress real noise blindly. Since denoising performance is closely related to the noise level, frame displacement, and the number of frames (burst length), intensive experiments including ablation study are performed. Quantitative results show that the proposed method performs significantly better than previous state‐of‐the‐art methods V‐BM4D and KPN in removing Gaussian noise. Qualitative results show that the proposed method is also effective in removing real noise using bursts and the SE module is key to reduce blur in results.

Image denoising aims to restore a clean image from a noisy image. Traditional methods utilizing convolutional neural networks (CNN) for denoising are trained using pairs of noisy and clean images to comprehend the transformation from a noisy image to a clean one. However, the acquisition of such image pairs in real‐world scenarios presents a challenge. Hence, numerous self‐supervised denoising techniques have been developed that do not require clean images for training. This study demonstrates that a straightforward loss design, concentrating on variance, can effectively train a standard CNN denoiser in a self‐supervised fashion. A novel theoretical framework is introduced for training a basic CNN denoising model using three constraints: mean, variance, and augmentation. The variance constraint is crucial as it prevents the trained model from converging to trivial solutions such as identity or zero mapping. This theory provides valuable insights for the development of new self‐supervised denoising methods. Furthermore, a method that applies this theory to proposed dual networks is developed, which consist of two standard CNN models predicting both the clean image and the noise. This approach enhances model capacity during training while minimizing computational costs during inference. This method exemplifies the implementation of the variance constraint and introduces a data constraint for dual networks. Notably, the proposed method only assumes the presence of additive white noise, irrespective of the noise distribution. This minimal assumption enhances the model's robustness against noise with complex or unknown distributions in real‐world distorted images. Experimental results indicate that the proposed Noise2Variance method exhibits commendable performance on peak signal noise ratio and structural similarity metrics compared to existing self‐supervised denoising techniques. Visual comparison of results further substantiates the efficacy of the proposed method. A comparison of model complexity reveals that the method is efficient among the compared CNN‐based techniques.

In the classical image processing pipeline, demosaicing and denoising are separated steps that may interfere with each other. Joint demosaicing and denoising utilizes the shared image prior information to guide the image recovery process. It is expected to have better performance by the joint optimization of the two problems. Besides, learning recovered images from burst (continuous exposure images) can further improve image details. This article proposes a two-stage convolutional neural network model for joint demosaicing and denoising of burst Bayer images. The proposed CNN model consists of a single-frame joint demosaicing and denoising module, a multiframe denoising module, and an optional noise estimation module. It requires a two-stage training scheme to ensure that the model converges to a good solution. Experiments on multiframe Bayer images with simulated Gaussian noise show that the proposed method has obvious performance advantages and speed advantages compared with similar approaches. Experiments on actual multiframe Bayer images verify the denoising effect and detail retention ability of the proposed method.