A Modified Tseng’s Method for Solving the Modified Variational Inclusion Problems and Its Applications

The goal of this study was to show how a modified variational inclusion problem can be solved based on Tseng’s method. In this study, we propose a modified Tseng’s method and increase the reliability of the proposed method. This method is to modify the relaxed inertial Tseng’s method by using certain conditions and the parallel technique. We also prove a weak convergence theorem under appropriate assumptions and some symmetry properties and then provide numerical experiments to demonstrate the convergence behavior of the proposed method. Moreover, the proposed method is used for image restoration technology, which takes a corrupt/noisy image and estimates the clean, original image. Finally, we show the signal-to-noise ratio (SNR) to guarantee image quality.

Particle Swarm Optimization and Fuzzy C-means for Domain-independent Noisy Image Segmentation

<p>Image segmentation is considered to be one of the foremost image analysis techniques for high-level real-world applications in computer vision. Its task is to change or simplify the representation of an image in order to make it easier to understand or analyze. Although image segmentation has been studied for many years, evolving technology and transformation of demands make image segmentation a continuing challenge. Noise as a side effect of imaging devices is an inevitable part of images in many computer vision applications. Therefore, an important topic in image segmentation is noisy image segmentation which requires extra effort to deal with image segmentation in the presence of noise. Generally, different strategies are needed for different noisy images with different levels/types of noise. Therefore, many approaches in the literature are domain-dependent and applicable only to specific images. A well-recognized approach in noisy image segmentation uses clustering algorithms, among which Fuzzy C-Means (FCM) is one of the most popular. FCM is unsupervised, efficient, and can deal with uncertainty and complexity of information in an image. Dealing with uncertainties is easier with the fuzzy characteristic of FCM, and complexity of information is being taken care of by utilizing different features in FCM, and also combining FCM with other techniques. Many modifications have been introduced to FCM to deal with noisy image segmentation more effectively. Common approaches include, adding spatial information into the FCM process, addressing the FCM initialization problem, and enhancing features used for segmentation. However, existing FCM-based noisy image segmentation approaches in the literature generally suffer from three drawbacks. First, they are applicable to specific domains and images, and impotent in others. Second, they don’t perform well on severely noisy image segmentation. Third, they are effective on specific type and level of noise, and they don’t explore the effect of noise level variation. Recently, evolutionary computation techniques due to their global search abilities have been used in hybridization with FCM, mostly to address FCM stagnation in local optima. Particle Swarm Optimization (PSO) is particularly of interest because of its lower computational costs, easy implementation, and fast convergence, but its potential in this area has not been fully investigated. This thesis develops new domain-independent PSO-based algorithms for an automatic non-supervised FCM-based segmentation of severely noisy images which are capable of extracting the main coherent/homogeneous regions while preserving details and being robust to noise variation. The key approach taken in the thesis is to explore the use of PSO to manipulate and enhance local spatial and spatial-frequency information. This thesis introduces a new PSO feature enhancement approach in wavelet domain for noisy image segmentation. This approach applies adaptive wavelet shrinkage using evaluation based on FCM clustering performance. The results show great accuracy in the case of severe noise because of the enhanced features. Also, due to adaptivity, no parameter-tuning is required according to the type or volume of noise, and the performance is consistent under noise level variation. This thesis presents a scheme under which a fusion of two different denoising algorithms for more effective segmentation is possible. This fusion retains the advantages of each algorithm while leaving out their drawbacks. The fusion scheme uses the noisy image segmentation system introduced above and anisotropic diffusion, the edge-preserving denoising algorithm. Results show greater accuracy and stability in comparison to the individual algorithms on a variety of noisy images. This thesis introduces another PSO-based edge-preserving adaptive wavelet shrinkage system using wavelet packets, bilateral filtering, and a detail-respecting shrinkage scheme. The analysis of the results provide a comparison between the two feature enhancement systems. The first system uses wavelets and the second uses wavelet packets as a domain to enhance features for an FCM-based noisy image segmentation. Also, the highest segmentation accuracy among all the algorithms introduced in this thesis on some benchmarks belong to this system.</p>

Particle Swarm Optimization and Fuzzy C-means for Domain-independent Noisy Image Segmentation

<p>Image segmentation is considered to be one of the foremost image analysis techniques for high-level real-world applications in computer vision. Its task is to change or simplify the representation of an image in order to make it easier to understand or analyze. Although image segmentation has been studied for many years, evolving technology and transformation of demands make image segmentation a continuing challenge. Noise as a side effect of imaging devices is an inevitable part of images in many computer vision applications. Therefore, an important topic in image segmentation is noisy image segmentation which requires extra effort to deal with image segmentation in the presence of noise. Generally, different strategies are needed for different noisy images with different levels/types of noise. Therefore, many approaches in the literature are domain-dependent and applicable only to specific images. A well-recognized approach in noisy image segmentation uses clustering algorithms, among which Fuzzy C-Means (FCM) is one of the most popular. FCM is unsupervised, efficient, and can deal with uncertainty and complexity of information in an image. Dealing with uncertainties is easier with the fuzzy characteristic of FCM, and complexity of information is being taken care of by utilizing different features in FCM, and also combining FCM with other techniques. Many modifications have been introduced to FCM to deal with noisy image segmentation more effectively. Common approaches include, adding spatial information into the FCM process, addressing the FCM initialization problem, and enhancing features used for segmentation. However, existing FCM-based noisy image segmentation approaches in the literature generally suffer from three drawbacks. First, they are applicable to specific domains and images, and impotent in others. Second, they don’t perform well on severely noisy image segmentation. Third, they are effective on specific type and level of noise, and they don’t explore the effect of noise level variation. Recently, evolutionary computation techniques due to their global search abilities have been used in hybridization with FCM, mostly to address FCM stagnation in local optima. Particle Swarm Optimization (PSO) is particularly of interest because of its lower computational costs, easy implementation, and fast convergence, but its potential in this area has not been fully investigated. This thesis develops new domain-independent PSO-based algorithms for an automatic non-supervised FCM-based segmentation of severely noisy images which are capable of extracting the main coherent/homogeneous regions while preserving details and being robust to noise variation. The key approach taken in the thesis is to explore the use of PSO to manipulate and enhance local spatial and spatial-frequency information. This thesis introduces a new PSO feature enhancement approach in wavelet domain for noisy image segmentation. This approach applies adaptive wavelet shrinkage using evaluation based on FCM clustering performance. The results show great accuracy in the case of severe noise because of the enhanced features. Also, due to adaptivity, no parameter-tuning is required according to the type or volume of noise, and the performance is consistent under noise level variation. This thesis presents a scheme under which a fusion of two different denoising algorithms for more effective segmentation is possible. This fusion retains the advantages of each algorithm while leaving out their drawbacks. The fusion scheme uses the noisy image segmentation system introduced above and anisotropic diffusion, the edge-preserving denoising algorithm. Results show greater accuracy and stability in comparison to the individual algorithms on a variety of noisy images. This thesis introduces another PSO-based edge-preserving adaptive wavelet shrinkage system using wavelet packets, bilateral filtering, and a detail-respecting shrinkage scheme. The analysis of the results provide a comparison between the two feature enhancement systems. The first system uses wavelets and the second uses wavelet packets as a domain to enhance features for an FCM-based noisy image segmentation. Also, the highest segmentation accuracy among all the algorithms introduced in this thesis on some benchmarks belong to this system.</p>

Self-Supervised Despeckling Algorithm with an Enhanced U-Net for Synthetic Aperture Radar Images

Self-supervised method has proven to be a suitable approach for despeckling on synthetic aperture radar (SAR) images. However, most self-supervised despeckling methods are trained by noisy-noisy image pairs, which are constructed by using natural images with simulated speckle noise, time-series real-world SAR images or generative adversarial network, limiting the practicability of these methods in real-world SAR images. Therefore, in this paper, a novel self-supervised despeckling algorithm with an enhanced U-Net is proposed for real-world SAR images. Firstly, unlike previous self-supervised despeckling works, the noisy-noisy image pairs are generated from real-word SAR images through a novel generation training pairs module, which makes it possible to train deep convolutional neural networks using real-world SAR images. Secondly, an enhanced U-Net is designed to improve the feature extraction and fusion capabilities of the network. Thirdly, a self-supervised training loss function with a regularization loss is proposed to address the difference of target pixel values between neighbors on the original SAR images. Finally, visual and quantitative experiments on simulated and real-world SAR images show that the proposed algorithm notably removes speckle noise with better preserving features, which exceed several state-of-the-art despeckling methods.

Noise2Grad: Extract Image Noise to Denoise

In many image denoising tasks, the difficulty of collecting noisy/clean image pairs limits the application of supervised CNNs. We consider such a case in which paired data and noise statistics are not accessible, but unpaired noisy and clean images are easy to collect. To form the necessary supervision, our strategy is to extract the noise from the noisy image to synthesize new data. To ease the interference of the image background, we use a noise removal module to aid noise extraction. The noise removal module first roughly removes noise from the noisy image, which is equivalent to excluding much background information. A noise approximation module can therefore easily extract a new noise map from the removed noise to match the gradient of the noisy input. This noise map is added to a random clean image to synthesize a new data pair, which is then fed back to the noise removal module to correct the noise removal process. These two modules cooperate to extract noise finely. After convergence, the noise removal module can remove noise without damaging other background details, so we use it as our final denoising network. Experiments show that the denoising performance of the proposed method is competitive with other supervised CNNs.