Noisy Input Image Research Articles

Improvement of noisy images filtered by bilateral process using a multi-scale context aggregation network

Deep learning has recently received a lot of attention as a feasible solution to a variety of artificial intelligence difficulties. Convolutional neural networks (CNNs) outperform other deep learning architectures in the application of object identification and recognition when compared to other machine learning methods. Speech recognition, pattern analysis, and image identification, all benefit from deep neural networks. When performing image operations on noisy images, such as fog removal or low light enhancement, image processing methods such as filtering or image enhancement are required. The study shows the effect of using Multi-scale deep learning Context Aggregation Network CAN on Bilateral Filtering Approximation (BFA) for de-noising noisy CCTV images. Data-store is used tomanage our dataset, which is an object or collection of data that are huge to enter in memory, it allows to read, manage, and process data located in multiple files as a single entity. The CAN architecture provides integral deep learning layers such as input, convolution, back normalization, and Leaky ReLu layers to construct multi-scale. It is also possible to add custom layers like adaptor normalization (µ) and adaptive normalization (Lambda) to the network. The performance of the developed CAN approximation operator on the bilateral filtering noisy image is proven when improving both the noisy reference image and a CCTV foggy image. The three image evaluation metrics (SSIM, NIQE, and PSNR) evaluate the developed CAN approximation visually and quantitatively when comparing the created de-noised image over the reference image.Compared with the input noisy image, these evaluation metrics for the developed CAN de-noised image were (0.92673/0.76253, 6.18105/12.1865, and 26.786/20.3254) respectively

Transmission reconstruction algorithm by combining maximum-likelihood expectation maximization and a convolutional neural network for radioactive drum characterization

One of the most critical factors that affect reconstructions of the activity of a radioactive drum is the accuracy of tomographic gamma scanning transmission reconstruction. The traditional algorithms applied for reconstructing the density map, such as maximum-likelihood expectation maximization (MLEM), algebraic reconstruction technique, or filtered back-projection, produce a grid distribution with severe grid artifacts and a high level of noise, which significantly degrade the detail of the transmission image, thereby increasing the reconstruction error for both the density map and the activity. Thus, we propose a novel algorithm for transmission reconstruction by combining MLEM and a deep convolutional neural network (CNN). The CNN is trained using a supervised learning approach with image pairs obtained from the drum's ground truth (target) and the result constructed using MLEM (input). Our experimental results indicate that the proposed reconstruction algorithm can significantly improve the spatial resolution while also removing grid artifacts. In addition, the algorithm is sufficiently robust when dealing with a noisy input image. When the input image contained 20% noise, the mean squared error of the output image decreased by 78.51% compared with the conventional reconstruction method, and the peak signal-to-noise ratio and structural similarity index measure for the output image improved by 81.19% and 71.74%, respectively. The new algorithm improves the accuracy of the radioactive drum density map reconstruction as well as decreasing the measurement time required. We consider that the proposed algorithm is an effective new method for use in the field of radioactive waste drum transmission image reconstruction.

Restoration of Tc-99m methyl diphosphonate bone scan image using Richardson-Lucy algorithm.

In this study, the optimal input parameters point spread function (PSF) and the number of iterations of the Richardson-Lucy algorithm were experimentally determined to restore Tc-99 m methyl diphosphonate (MDP) whole-body bone scan images. The experiment was performed on 60 anonymized Tc-99 m MDP whole-body bone scan images. Ten images were used for estimating the optimum value of PSF and the number of iterations to restore scintigraphic images. The remaining 50 images were used for validation of estimated parameters. The image quality of observed and restored images was assessed objectively using blind/referenceless image spatial quality evaluator (BRISQUE), mean brightness (MB), discrete entropy (DE), and edge-based contrast measure (EBCM) image quality metrics. Image quality was subjectively assessed by two nuclear medicine physicians (NMPs) by comparing the restored image quality with observed image quality and assigning a score to each image on the scale of 0-5. Based on BRISQUE, MB, DE, and EBCM scores, the restored images were significantly sharper, less bright, had more detailed information, and had less contrast around edges compared to the input images. The restored images had improved resolution based on visual assessment as well; NMPs assigned an average image quality score of 4.00 to restored images. Maximum resolution enhancement was noticed at PSF (size: 11 pixels, sigma: 1.75 pixels) and the number of iterations = 10. With the increase in the number of iterations, noise also gets amplified along with resolution enhancement and affects the detectability of small lesions; in the case of relatively low noisy input images, the number of iterations = 5 gave better results. Tc-99 m MDP bone scan images were restored to improve image quality using the Richardson-Lucy algorithm. The optimum value of the PSF parameter was found to be of size = 11 pixels and sigma = 1.75 pixels.

Improving the performance of automotive vision‐based applications under rainy conditions

Input images are the main source of information for vision-based algorithms. The presence of raindrops in input images degrades their quality and, consequently, reduces the quality of the target vision-based algorithm that consumes them. Many image restoration algorithms were proposed in the literature to remove rain presence in images to improve the input image quality. These algorithms, however, cannot remove all the raindrop presence and sometimes introduce undesirable side-effects, such as the blurring rain-occluded sections of the image and incorrectly de-raining areas in the image that are clear. It is hypothesized that a comparable performance improvement can be achieved by decreasing the sensitivity of vision-based algorithms to noisy input images, rather than denoising these images, through the process of de-raining. To test this hypothesis, the performance of state-of-the-art object detection and semantic segmentation models was evaluated, with de-rained image datasets used as input, and compared it to that performance of the same models, retrained with rained image sets. Results showed that the performance of the retrained models was better than that of the baseline detector with de-rained images used as input.

A novel framework for segmentation of uterus fibroids in ultrasound images using machine learning models

A tumour of non-cancerous structure that appears in uterus during child-bearing years is uterine fibroids. Thus, it is necessary to design a fibroid detection system for the fibroid ablation. Various methods developed for the detection of fibroids are easily affected by the image artefacts as they do not take into consideration the spatial information and have lower efficiency problems for fibroid segmentation. This paper puts forward a method for segmentation for fibroid detection. The proposed segmentation model overcomes the drawbacks of existing methodologies of fibroid detection in all stages. Here, the speckle noise existing in the noisy input image can be removed by using IGDT-DWT method and EMD-GCLAHE method. After contrast enhancement, the segmentation of the contrast-enhanced image is done using a novel clustering algorithm namely PC-K-mean algorithm. The proposed segmentation algorithm effectively detects the fibroids, which is experimentally proved by comparing it with existing classifiers.

Multilevel Noise Contrastive Network for Few-Shot Image Denoising

In recent years, most denoising methods based on deep Convolutional Neural Networks heavily rely on massive noisy-clean image pairs. Collecting massive noisy-clean image pairs is expensive and not practical in real scenes. Currently, few-shot learning has been applied to many areas to cope with the absence data. The few-shot learning, however, in image denoising severely suffers from domain gap problems, including dataset domain gap and feature domain gap, especially for the real noisy images. Therefore, this paper proposes a Multi-level Noise Contrastive Network (MNC-Net) performing few-shot image denoising. MNC-Net consists of two training stages: i) using contrastive learning to self-supervise the training of Multi-level Noise Contrastive Learner (MNCL) on the pure synthetic noisy images with multiple Gaussian noise levels to ease the acute dataset domain gap, and ii) features generated by the MNCL on limited data are fused to the second stage and alleviate the feature domain gap using our proposed denoising network. Specifically, the MNCL consists of a Contrastive Feature Extractor (CFE) and a Contrastive Feature Projector (CFP). MNCL learns the rich and complex content-invariant degradations and general multiple-level noise representations. The denoising network in the second stage is composed of Guider Feature Encoder (GFE) and Adaptive Denoising Decoder (ADD). The GFE uses contrast features from CFE to guide the produced representations on the specific input noisy images. Then, such output features are fed into the ADD to adaptively denoise the noisy images on the corresponding noise distribution. To our best knowledge, this work is the first attempt to jointly use the few-shot learning and contrastive learning in the deep denoising field. Extensive experiments on CBSD68, Kodak24, Set12, SIDD, and DND show that our method achieves promising denoising performances in the absence of data.

Texture Feature Extraction of Lumbar Spine Trabecular Bone Radiograph Image using Laplacian of Gaussian Filter with KNN Classification to Diagnose Osteoporosis

The human bones are categorized based on elemental micro architecture and porosity. The porosity of the inner trabecular bone is high that is 40-95% and the nature of the bone is soft and spongy whereas the cortical bone is harder and is less porous that is 5 to 15%. Osteoporosis is a disease that normally affects women usually after their menopause. It largely causes mild bone fractures and further stages lead to the demise of an individual. The detection of Osteoporosis in Lumbar Spine has been widely recognized as a promising way to frequent fractures. Therefore, premature analysis of osteoporosis will estimate the risk of the bone fracture which prevents life threats. The paper is systematized in two different sections to classify normal (non-osteoporosis) and abnormal(osteoporosis)Lumbar spine trabecular bone. In this method, the first section is based on discriminating the lumbar spine trabecular bone micro-architecture predisposing by means of first and second order directional derivative of Laplacian of Gaussian filter with different standard deviation to acquire the minimum and maximum responses. The dimension reduction of texture features, quantization and adjacent scale coding with weighted multipliers are used to lessen the intensity variations of texture features. The second section is based on the reduction of histogram features as a training data set for classification of normal and osteoporotic images of lumbar spine (L1-L4) using K-Nearest Neighborhood (KNN) classifier. The tested dataset result gives effective classification accuracy of 97.22% with lesser texture feature dimension. The usage of weight multiplier as well as quantization technique plays a major role for the improvement of accuracy to diagnose osteoporosis for an input noisy and noiseless image.

Strided Self-Supervised Low-Dose CT Denoising for Lung Nodule Classification.

Lung nodule classification based on low-dose computed tomography (LDCT) images has attracted major attention thanks to the reduced radiation dose and its potential for early diagnosis of lung cancer from LDCT-based lung cancer screening. However, LDCT images suffer from severe noise, largely influencing the performance of lung nodule classification. Current methods combining denoising and classification tasks typically require the corresponding normal-dose CT (NDCT) images as the supervision for the denoising task, which is impractical in the context of clinical diagnosis using LDCT. To jointly train these two tasks in a unified framework without the NDCT images, this paper introduces a novel self-supervised method, termed strided Noise2Neighbors or SN2N, for blind medical image denoising and lung nodule classification, where the supervision is generated from noisy input images. More specifically, the proposed SN2N can construct the supervision information from its neighbors for LDCT denoising, which does not need NDCT images anymore. The proposed SN2N method enables joint training of LDCT denoising and lung nodule classification tasks by using self-supervised loss for denoising and cross-entropy loss for classification. Extensively experimental results on the Mayo LDCT dataset demonstrate that our SN2N achieves competitive performance compared with the supervised learning methods that have paired NDCT images as supervision. Moreover, our results on the LIDC-IDRI dataset show that the joint training of LDCT denoising and lung nodule classification significantly improves the performance of LDCT-based lung nodule classification.

Adaptive Removal of high-density salt-and-pepper noise (ARSPN) for robust ROI detection used in watermarking of MRI images of the brain

This paper presents a novel window-based method to remove high-density salt-and-pepper noise for optimal ROI (Region Of Interest) detection of brain MRI (Magnetic Resonance Imaging) images. The output of this system is used in watermarking and extracting hidden data in this type of image. In this method, for each pixel of the noisy input image, an adaptive n × n window is considered in the neighborhood of that pixel. If they are not noisy, the pixels of this window are weighted according to their distance from the desired pixel, and the weighted sum of the neighboring pixels is averaged. Then the noisy pixel replaces with the resulting value. This paper consists of three main sections: ROI detection, noise removal block, and evaluation of the proposed method against different densities of salt-and-pepper noise in the range of 1%–98%. ROI obtained by this method is the same before and after the noise. The final image has an acceptable PSNR (Peak Signal-to-Noise Ratio) for noise with various densities. Based on the experimental results obtained by the high efficient proposed noise removal method using 208 images from seven Databases (DBs), the maximum value is 61.7% for the 1% noise density and 26.4% for the 98% noise density.

An Advanced Noise Reduction and Edge Enhancement Algorithm

Complementary metal-oxide-semiconductor (CMOS) image sensors can cause noise in images collected or transmitted in unfavorable environments, especially low-illumination scenarios. Numerous approaches have been developed to solve the problem of image noise removal. However, producing natural and high-quality denoised images remains a crucial challenge. To meet this challenge, we introduce a novel approach for image denoising with the following three main contributions. First, we devise a deep image prior-based module that can produce a noise-reduced image as well as a contrast-enhanced denoised one from a noisy input image. Second, the produced images are passed through a proposed image fusion (IF) module based on Laplacian pyramid decomposition to combine them and prevent noise amplification and color shift. Finally, we introduce a progressive refinement (PR) module, which adopts the summed-area tables to take advantage of spatially correlated information for edge and image quality enhancement. Qualitative and quantitative evaluations demonstrate the efficiency, superiority, and robustness of our proposed method.

Image Denoising Using Adaptive and Overlapped Average Filtering and Mixed-Pooling Attention Refinement Networks

Cameras are essential parts of portable devices, such as smartphones and tablets. Most people have a smartphone and can take pictures anywhere and anytime to record their lives. However, these pictures captured by cameras may suffer from noise contamination, causing issues for subsequent image analysis, such as image recognition, object tracking, and classification of an object in the image. This paper develops an effective combinational denoising framework based on the proposed Adaptive and Overlapped Average Filtering (AOAF) and Mixed-pooling Attention Refinement Networks (MARNs). First, we apply AOAF to the noisy input image to obtain a preliminarily denoised result, where noisy pixels are removed and recovered. Next, MARNs take the preliminary result as the input and output a refined image where details and edges are better reconstructed. The experimental results demonstrate that our method performs favorably against state-of-the-art denoising methods.

Complete and representative training of neural networks: A generalization study using double noise injection and natural images

Neural networks hold substantial promise for automating various processing and interpretation tasks. Yet, their performance is often suboptimal compared with standard but more closely guided approaches. The lack of performance is often attributed to poor generalization, in particular if fewer training examples are provided than free parameters that exist in the machine-learning algorithm. In this case, the training data are typically memorized instead of the algorithm learning the underlying general trends. The network generalization is improved if the provided samples are representative, in that they describe all features of interest well. We argue that a more subtle condition preventing poor performance is that the provided examples must also be complete; the examples must span the full solution space. Ensuring completeness during training is challenging unless the target application is well understood. We illustrate that one possible solution is to make the problem more general if this greatly increases the number of available training data. For instance, if seismic images are treated as a subclass of natural images, then a deep-learning-based denoiser for seismic data can be trained using exclusively natural images. The latter are widely available. The resulting denoising algorithm has never seen any seismic data during the training stage, yet it displays performance comparable with standard and advanced random-noise reduction methods. We exclude any seismic data during training to demonstrate that natural images are complete and representative for this specific task. Furthermore, we apply a novel approach to increase the amount of training data known as double noise injection, providing noisy input and output images during the training process. Given the importance of network generalization, we hope that the insights that we gained may help improve the performance of a range of machine-learning applications in geophysics.

Image Impulse Noise Removal Using Cascaded Filtering Based on Overlapped Adaptive Gaussian Smoothing and Convolutional Refinement Networks

Impulse noise is often introduced to images when captured through image sensors due to sharp and sudden disturbances in the image signal, analog-to-digital converter errors, sensor temperature, etc., severely degrading their visual quality. Therefore, it is essential to develop an effective method to remove image noise. We propose a novel image denoising method for “salt-and-pepper” (SP) noise, using cascaded filtering based on overlapped adaptive Gaussian smoothing (OAGS) and the convolutional refinement networks (CRNs). First, the noisy input image can be preliminarily denoised by OAGS, where the noisy pixels are removed and recovered. The CRNs refine the result by restoring fine details for the denoised image. Through extensive experimental results, we demonstrate the proposed method substantially outperforms other state-of-the-art denoising methods, especially for high-density SP noise.

Hybrid Between Ontology and Quantum Particle Swarm Optimization for Segmenting Noisy Plant Disease Image

One of the main risks to food security is plant diseases, but because of the absence of needed infrastructure and actual noise, scientists are faced with a difficult issue. Semantic segmentation of images divides images into non-overlapped regions, with specified semantic labels allocated. In this paper, The QPSO (quantum particle swarm optimization) algorithm has been used in segmentation of an original noisy image and Ontology has been used in classification the segmented image. Input noisy image segmentation is limited to a classification phase in which the object is transferred to Ontology. With 49,563 images from healthy and diseased plant leaves, 12 plant species were identified and 22 diseases, the proposed method is evaluated. The method proposed produces an accuracy of 86.22 percent for a stopped test set, showing that the strategy is appropriate. EPDO (Enhance Plant Disease Ontology) is built with the web ontology language (OWL). The segmented noisy image elements are paired with EPDO with derived features that come from QPSO. Our results show that a classification based on the suggested method is better than the state-of-the-art algorithms. The proposed method also saves time and effort for removing the noise at noise level from the input image σ=70

Learning to denoise astronomical images with U-nets

ABSTRACT Astronomical images are essential for exploring and understanding the Universe. Optical telescopes capable of deep observations, such as the Hubble Space Telescope (HST), are heavily oversubscribed in the Astronomical Community. Images also often contain additive noise, which makes denoising a mandatory step in post-processing the data before further data analysis. In order to maximize the efficiency and information gain in the post-processing of astronomical imaging, we turn to machine learning. We propose Astro U-net, a convolutional neural network for image denoising and enhancement. For a proof-of-concept, we use HST images from Wide Field Camera 3 instrument UV/visible channel with F555W and F606W filters. Our network is able to produce images with noise characteristics as if they are obtained with twice the exposure time, and with minimum bias or information loss. From these images, we are able to recover $95.9{{\ \rm per\ cent}}$ of stars with an average flux error of $2.26{{\ \rm per\ cent}}$. Furthermore, the images have, on average, 1.63 times higher signal-to-noise ratio than the input noisy images, equivalent to the stacking of at least three input images, which means a significant reduction in the telescope time needed for future astronomical imaging campaigns.

Additive White Gaussian Noise Level Estimation for Natural Images Using Linear Scale-Space Features

Noise in images is often modelled with additive white Gaussian noise (AWGN). An accurate estimation of noise level without any prior knowledge of noisy input image leads to effective blind image denoising methods. The performance of certain image denoising methods under AWGN model is dependent on the accuracy of noise level estimation (NLE). Hence, there is a need to develop an effective NLE method in order to achieve better performance in image denoising. Even though the existing NLE methods perform well on natural images, these methods involve complex segmentation tasks such as homogeneous regions extraction and super-pixel decomposition. Hence, a simple, fast, and accurate NLE method for AWGN is proposed in this paper. In the presented NLE method, the statistical features of high-frequency details of noisy input image are obtained at multiple linear (Gaussian) scale-space which are used to construct a feature vector. It is perceived that the features obtained are almost linear and separable. Hence, supervised linear regression (LR) models that are trained globally and locally are suggested for NLE. The proposed method estimates the noise level in two stages. In stage-1, a globally trained LR model is used to estimate the noise level. It is observed that the accuracy of the noise level obtained through stage-1 can be further improved in stage-2 by adopting the proposed locally trained LR model. The proposed NLE method is evaluated with artificially generated noisy natural images using AWGN model. The high-quality natural images from Waterloo and BSD500 datasets are selected using image quality selection module and then used in training and testing phases. The average absolute deviation (AAD) is evaluated from each selected image in the datasets over a wide range of noise levels ([0 100]). The average AAD for selected images in Waterloo (BSD500) dataset is 0.21 (0.18), and execution time required to estimate the noise level is 0.04 s per image. From the obtained results, it is clear that the proposed method is simple, fast, and accurate as compared to several existing NLE methods. The effectiveness of the proposed NLE method is illustrated with fast and flexible denoising convolutional neural network using standard test images at randomly selected noise levels.

A new image denoising framework using bilateral filtering based non-subsampled shearlet transform

In this paper, we propose an advanced framework for image denoising using bilateral filtering based non-subsampled shearlet transform (NSST). Initially, we apply the NSST to decompose noisy input image into high and low frequency coefficients. The weighted bilateral filter (WBF) is then used to remove the noise from the low frequency coefficients while; thresholding is used to remove noise from the high frequency coefficients. The outputs of both the process are combined to form the resultant image. Finally, the inverse NSST is applied on the resultant output to estimate the final denoised image. To ensure validity of the proposed model, we conduct several experiments by considering different grayscale images with various noise variances. The qualitative and quantitative comparison is illustrated and it shows the improved performance of the proposed method as compared to the other conventional image denoising methods. Mathematical and simulation results are presented to show the validity of our work.

End-to-End Unpaired Image Denoising with Conditional Adversarial Networks

Image denoising is a classic low level vision problem that attempts to recover a noise-free image from a noisy observation. Recent advances in deep neural networks have outperformed traditional prior based methods for image denoising. However, the existing methods either require paired noisy and clean images for training or impose certain assumptions on the noise distribution and data types. In this paper, we present an end-to-end unpaired image denoising framework (UIDNet) that denoises images with only unpaired clean and noisy training images. The critical component of our model is a noise learning module based on a conditional Generative Adversarial Network (cGAN). The model learns the noise distribution from the input noisy images and uses it to transform the input clean images to noisy ones without any assumption on the noise distribution and data types. This process results in pairs of clean and pseudo-noisy images. Such pairs are then used to train another denoising network similar to the existing denoising methods based on paired images. The noise learning and denoising components are integrated together so that they can be trained end-to-end. Extensive experimental evaluation has been performed on both synthetic and real data including real photographs and computer tomography (CT) images. The results demonstrate that our model outperforms the previous models trained on unpaired images as well as the state-of-the-art methods based on paired training data when proper training pairs are unavailable.

Hybrid image denoising based on region division

This paper proposes an efficient hybrid framework for image denoising, in which the advantages of different denoising methods are effectively incorporated by using the region prior knowledge. In detail, the input noisy image is first divided into a large number of overlapping patches followed by extraction of Speed Up Robust Feature (SURF), and then the noisy patches are classified into two categories based on Twin Support Vector Machine (TWSVM). The texture patches are enhanced via Gradient Histogram Preservation (GHP) while flat patches can be re-established using Block Matching Three-Dimensional Filtering (BM3D). Finally, the re-established images can be acquired by fusing the processing results of the two kinds of patches. To evaluate the effectiveness of the presented method, we conduct experiments on standard image datasets and compare the performance with other outstanding denoising approaches. Experimental results show that the presented method achieves better results, especially in containing textures and edges compared with existing image denoising methods.

Hybrid image denoising based on region division

This paper proposes an efficient hybrid framework for image denoising, in which the advantages of different denoising methods are effectively incorporated by using the region prior knowledge. In detail, the input noisy image is first divided into a large number of overlapping patches followed by extraction of Speed Up Robust Feature (SURF), and then the noisy patches are classified into two categories based on Twin Support Vector Machine (TWSVM). The texture patches are enhanced via Gradient Histogram Preservation (GHP) while flat patches can be re-established using Block Matching Three-Dimensional Filtering (BM3D). Finally, the re-established images can be acquired by fusing the processing results of the two kinds of patches. To evaluate the effectiveness of the presented method, we conduct experiments on standard image datasets and compare the performance with other outstanding denoising approaches. Experimental results show that the presented method achieves better results, especially in containing textures and edges compared with existing image denoising methods.