This paper proposes a novel framework for enhancing underwater images captured by optical imaging model and non-local means denoising. The proposed approach adjusts the color balance using biasness correction and the average luminance. Scene visibility is then enhanced based on an underwater optical imaging model. The increase in noise in the enhanced images is alleviated by non-local means (NLM) denoising. The final enhanced images are characterized by improved visibility while retaining color fidelity and reducing noise. The proposed method does not require specialized hardware nor prior knowledge of the underwater environment.

An improved multivariate wavelet denoising algorithm combined with subspace and principal component analysis is presented in this paper. The key element is deriving an optimal orthogonal matrix that can project the multivariate observation signal to a signal subspace from observation space. Univariate wavelet shrinkage operator is then applied to the projected signals channel-wise resulting in the improvement of the output SNR. Finally, principal component analysis is performed on the denoised signal in the observation space to further improve the denoising performance. Experimental results based on synthesized and real world ECG data verify the effectiveness of the proposed algorithm.

This letter introduces a non-local means (NLM) denoising algorithm that uses a weight function based on a switching norm. The noise level and local activity are incorporated into the NLM denoising algorithm which enhances performance. This is done by selecting a norm among l1, l2, and l4 norms to determine a weighting function. The experimental results show the capability of the proposed algorithm. In addition, the proposed algorithm is verified as effective for enhancing the performance of other NLM algorithms.

We propose an image restoration technique that uses multiple image integration. The detail of the dark area when acquiring a dark scene is often deteriorated by sensor noise. Simple image integration inherently has the capability of reducing random noises, but it is especially insufficient in scenes that have a dark area. We introduce a novel image integration technique that optimizes the weights for the integration. We find the optimal weight map by solving a convex optimization problem for the weight optimization. Additionally, we apply the proposed weight optimization scheme to a single-image super-resolution problem, where we slightly modify the weight optimization problem to estimate the high-resolution image from a single low-resolution one. We use some of our experimental results to show that the weight optimization significantly improves the denoising and super-resolution performances.

In this letter, we propose a novel supervised pre-training technique for deep neural network (DNN)-hidden Markov model systems to achieve robust speech recognition in adverse environments. In the proposed approach, our aim is to initialize the DNN parameters such that they yield abstract features robust to acoustic environment variations. In order to achieve this, we first derive the abstract features from an early fine-tuned DNN model which is trained based on a clean speech database. By using the derived abstract features as the target values, the standard error back-propagation algorithm with the stochastic gradient descent method is performed to estimate the initial parameters of the DNN. The performance of the proposed algorithm was evaluated on Aurora-4 DB, and better results were observed compared to a number of conventional pre-training methods.

This paper introduces a fast image denoising algorithm by estimating noise parameters without prior information about the noise. Under the assumption that additive noise has a Gaussian distribution, the noise parameters were estimated from an observed degraded image, and were used to define the constraints of a noise detection process that was coupled with a Markov random field (MRF). In addition, an adaptive modified weighted Gaussian filter with variable window sizes defined by the constraints on noise detection was used to control the degree of smoothness of the reconstructed image. Experimental results demonstrate the capability of the proposed algorithm.

A cascaded video denoising method based on frame averaging is proposed in this paper. A novel segmentation approach using intensity and structure tensor is used for change compensation, which can effectively suppress noise while preserving the structure of an image. The cascaded framework solves the problem of noise residual caused by single-frame averaging. The classical Wiener filter is used for spatial denoising in changing areas. Our algorithm works in real-time on an FPGA, since it does not involve future frames. Experiments on standard grayscale videos for various noise levels demonstrate that the proposed method is competitive with current state-of-the-art video denoising algorithms on both peak signal-to-noise ratio and structural similarity evaluations, particularly when dealing with large-scale noise.

Using traditional single-layer dictionary learning methods, it is difficult to reveal the complex structures hidden in the hyperspectral images. Motivated by deep learning technique, a deep dictionary learning approach is proposed for hyperspectral image denoising, which consists of hierarchical dictionary learning, feature denoising and fine-tuning. Hierarchical dictionary learning is helpful for uncovering the hidden factors in the spectral dimension, and fine-tuning is beneficial for preserving the spectral structure. Experiments demonstrate the effectiveness of the proposed approach.

In this paper, an image prior based on soft-morphological filters and its application to image recovery are presented. In morphological image processing, a gray-scale image is represented as a subset in a three-dimensional space, which is spanned by spatial and intensity axes. Morphological opening and closing, which are basic operations in morphological image processing, respectively approximate the image subset and its complementary images as the unions of structuring elements that are translated in the three-dimensional space. In this study, the opening and closing filters are applied to an image prior to resolve the regularization problem of image recovery. When the proposed image prior is applied, the image is recovered as an image that has no noise component, which is eliminated by the opening and closing. However, the closing and opening filters are less able to eliminate Gaussian noise. In order to improve the robustness against Gaussian noise, the closing and opening filters are respectively approximated as soft-closing and soft-opening with relaxed max and min functions. In image recovery experiments, image denoising and deblurring using the proposed prior are demonstrated. Comparisons of the proposed prior with the existing priors that impose a penalty on the gradient of the intensity are also shown.

In this letter, we present an efficient video matching-based denoising method. Two main issues are addressed in this paper: the matched points and the denoising algorithm based on an adaptive spatial temporal filter. Unlike previous algorithms, our method adaptively selects reference pixels within spatially and temporally neighboring frames. Our method uses more information about matched pixels on neighboring frames than other methods. Therefore, the proposal enhanced the accuracy of video denoising. Simulation results show that the proposed method produces cleaner and sharper images.

In this paper, we propose a novel technique for film grain noise removal and synthesis that can be adopted in high fidelity video coding. Film grain noise enhances the natural appearance of high fidelity video, therefore, it should be preserved. However, film grain noise is a burden to typical video compression systems because it has relatively large energy levels in the high frequency region. In order to improve the coding performance while preserving film grain noise, we propose film grain noise removal in the pre-processing step and film grain noise synthesis in the post processing step. In the pre-processing step, the film grain noise is removed by using temporal and inter-color correlations. Specifically, color image denoisng using inter color prediction provides good denoising performance in the noise-concentrated B plane, because film grain noise has inter-color correlation in the RGB domain. In the post-processing step, we present a noise model to generate noise that is close to the actual noise in terms of a couple of observed statistical properties, such as the inter-color correlation and power of the film grain noise. The results show that the coding gain of the denoised video is higher than for previous works, while the visual quality of the final reconstructed video is well preserved.

In this paper, we propose a novel multi-frame image denoising technique, which achieves the minimum variance of noise. Zero-mean and unknown variance white noise with an arbitrary distribution is considered in this paper. The proposed method consists of two parts. The first one is the estimation of the variance of noise for each image by considering the differences of all pairs of images. The second one is an actual denoising process in which the convex combination of all images with weight coefficients determined by the estimated variances is constructed. We also give an efficient algorithm by which we can obtain the same result by successive convex combinations. The efficacy of the proposed method is confirmed by computer simulations.

We propose a maximum likelihood estimation approach for the recovery of continuously-defined sparse signals from noisy measurements, in particular periodic sequences of Diracs, derivatives of Diracs and piecewise polynomials. The conventional approach for this problem is based on least-squares (a.k.a. annihilating filter method) and Cadzow denoising. It requires more measurements than the number of unknown parameters and mistakenly splits the derivatives of Diracs into several Diracs at different positions. Moreover, Cadzow denoising does not guarantee any optimality. The proposed approach based on maximum likelihood estimation solves all of these problems. Since the corresponding log-likelihood function is non-convex, we exploit the stochastic method called particle swarm optimization (PSO) to find the global solution. Simulation results confirm the effectiveness of the proposed approach, for a reasonable computational cost.

This paper proposes an adaptation method for structuring elements of morphological filters. A structuring element of a morphological filter specifies a shape of local structures that is eliminated or preserved in the output. The adaptation of the structuring element is hence a crucial problem for image denoising using morphological filters. Existing adaptation methods for structuring elements require preliminary training using example images. We propose an adaptation method for structuring elements of morphological opening filters that does not require such training. In our approach, the opening filter is interpreted as an approximation method with the union of the structuring elements. In order to eliminate noise components, a penalty defined from an assumption of image smoothness is imposed on the structuring element. Image denoising is achieved through decreasing the objective function, which is the sum of an approximation error term and the penalty function. In experiments, we use the proposed method to demonstrate positive impulsive noise reduction from images.

The noise in digital images acquired by image sensors has complex characteristics due to the variety of noise sources. However, most noise reduction methods assume that an image has additive white Gaussian noise (AWGN) with a constant standard deviation, and thus such methods are not effective for use with image signal processors (ISPs). To efficiently reduce the noise in an ISP, we estimate a unified noise model for an image sensor that can handle shot noise, dark-current noise, and fixed-pattern noise (FPN) together, and then we adaptively reduce the image noise using an adaptive Smallest Univalue Segment Assimilating Nucleus ( SUSAN ) filter based on the unified noise model. Since our noise model is affected only by image sensor gain, the parameters for our noise model do not need to be re-configured depending on the contents of image. Therefore, the proposed noise model is suitable for use in an ISP. Our experimental results indicate that the proposed method reduces image sensor noise efficiently.

Automatic modulation classification (AMC) involves extracting a set of unique features from the received signal. Accuracy and uniqueness of the features along with the appropriate classification algorithm determine the overall performance of AMC systems. Accuracy of any modulation feature is usually limited by the blindness of the signal information such as carrier frequency, symbol rate etc. Most papers do not sufficiently consider these impairments and so do not directly target practical applications. The AMC system proposed herein is trained with probable input signals, and the appropriate decision tree should be chosen to achieve robust classification. Six unique features are used to classify eight analog and digital modulation schemes which are widely used by low frequency mobile emergency radios around the globe. The Proposed algorithm improves the classification performance of AMC especially for the low SNR regime.

As a basic tool for deriving sparse representation of a color image from its atomic-decomposition with a redundant dictionary, the authors have recently proposed a new kind of shrinkage technique, viz. color shrinkage, which utilizes inter-channel color dependence directly in the three primary color space. Among various schemes of color shrinkage, this paper particularly presents the soft color-shrinkage and the hard color-shrinkage, natural extensions of the classic soft-shrinkage and the classic hard-shrinkage respectively, and shows their advantages over the existing shrinkage approaches where the classic shrinkage techniques are applied after a color transformation such as the opponent color transformation. Moreover, this paper presents the applications of our color-shrinkage schemes to color-image processing in the redundant tight-frame transform domain, and shows their superiority over the existing shrinkage approaches.

We propose, in this letter, a new type of image denoising filter using a data analysis technique. We deal with pixels as data and extract the most dominant cluster from pixels in the filtering window. We output the centroid of the extracted cluster. We demonstrate that this graph-spectral filter can effectively reduce a mixture of Gaussian and random impulsive noise.

The influence of noise is an important problem on image acquisition and transmission stages. The traditional image denoising approaches only analyzing the pixels of local region with a moving window, which calculated by neighbor pixels to denoise. Recently, this research has been focused on the transform domain and feature space. Compare with the traditional approaches, the global multi-scale analyzing and unchangeable noise distribution is the advantage. Apparently, the estimation based methods can be used in transform domain and get better effect. This paper proposed a new approach to image denoising in orthonormal wavelet domain. In this paper, we adopt Stein's unbiased risk estimate (SURE) based method to denoise the low-frequency bands and the feature patches distance constraint (FPDC) method also be proposed to estimate the noise free bands in Wavelet domain. The key point is that how to divide the lower frequency sub-bands and the higher frequency sub-bands, and do interscale SURE and intrascale FPDC, respectively. We compared our denoising method with some well-known and new denoising algorithms, the experimental results show that the proposed method can give better performance and keep more detail information in most objective and subjective criteria than other methods.

This letter proposes a novel design approach for optimal contourlet filter banks based on the parametric 9/7 filter family. The Laplacian pyramid decomposition is replaced by optimal 9/7 filter banks with rational coefficients, and directional filter banks are activated using a pkva 12 filter in the contourlets. Moreover, based on this optimal 9/7 filter, we present an image denoising approach using a contourlet domain hidden Markov tree model. Finally, experimental results show that our approach in denoising images with texture detail is only 0.20 dB less compared to the method of Po and Do, and the visual quality is as good as for their method. Compared with the method of Po and Do, our approach has lower computational complexity and is more suitable for VLSI hardware implementation.