Natural Grayscale Images Research Articles

Qutrit-Inspired Fully Self-Supervised Shallow Quantum Learning Network for Brain Tumor Segmentation.

Classical self-supervised networks suffer from convergence problems and reduced segmentation accuracy due to forceful termination. Qubits or bilevel quantum bits often describe quantum neural network models. In this article, a novel self-supervised shallow learning network model exploiting the sophisticated three-level qutrit-inspired quantum information system, referred to as quantum fully self-supervised neural network (QFS-Net), is presented for automated segmentation of brain magnetic resonance (MR) images. The QFS-Net model comprises a trinity of a layered structure of qutrits interconnected through parametric Hadamard gates using an eight-connected second-order neighborhood-based topology. The nonlinear transformation of the qutrit states allows the underlying quantum neural network model to encode the quantum states, thereby enabling a faster self-organized counterpropagation of these states between the layers without supervision. The suggested QFS-Net model is tailored and extensively validated on the Cancer Imaging Archive (TCIA) dataset collected from the Nature repository. The experimental results are also compared with state-of-the-art supervised (U-Net and URes-Net architectures) and the self-supervised QIS-Net model and its classical counterpart. Results shed promising segmented outcomes in detecting tumors in terms of dice similarity and accuracy with minimum human intervention and computational resources. The proposed QFS-Net is also investigated on natural gray-scale images from the Berkeley segmentation dataset and yields promising outcomes in segmentation, thereby demonstrating the robustness of the QFS-Net model.

Division of images into areas of local extrema with a monotonic change in pixel brightness

In this paper, the problem of segmentation of halftone images is considered, in which areas of local maxima and minima (extrema) are distinguished with a monotonic change in the brightness of pixels from local extrema to the boundaries of areas. To solve this problem, a mathematical model is proposed and a segmentation algorithm is developed on the basis of counter-wave growing of local extremum regions. The developed algorithm differs from the known segmentation algorithms by using a set of brightness thresholds (by the number of regions), varying by one in each cycle, starting from the values of local extrema, taking into account the increase or decrease in brightness to select adjacent pixels that are attached to the regions formed from these local extrema. The algorithm provides a greater deviation of pixel brightness from the average value within the region compared to known segmentation algorithms. This does not allow evaluating its efficiency using known indicators based on the variance of the brightness within the region. In this regard, estimates of the monotonicity of changes in the brightness of regions are proposed based on a) the shortest distances from each pixel of the region to the corresponding local extremum along the routes determined by the maximum increase (for the region of the local maximum) or decrease (for the region of the local minimum) the brightness of pixels and b) taking into account the number pixels that break the monotony of the segment brightness change. Using these estimates, it is shown that the proposed algorithm provides segmentation of artificial and natural grayscale images with a monotonic change in the brightness of pixels in the areas of local extrema. These properties allow us to consider the developed algorithm as a basis for the selection of texels, spots, low-contrast objects in images.

A spatially constrained asymmetric Gaussian mixture model for image segmentation

The finite Gaussian mixture model (GMM) is a flexible and powerful tool for addressing many computer vision and pattern recognition problems. The Gaussian distribution is a probability distribution that is symmetric with respect to the mean. However, in many segmentation applications, the observed data obey an asymmetric distribution. Furthermore, the GMM is sensitive to imaging noise. To alleviate these issues, a new finite anisotropic asymmetric normal mixture model is presented in this paper. Note that GMM is a degraded case of our proposed model. First, the proposed model employs anisotropic spatial information to reduce the effect of imaging noise while preserving the details, such as edges, corners and slim structure objects. Second, the anisotropic spatial information is coupled into the skew normal distribution to fit the observed data obeying an asymmetric distribution. Then the modeling and estimation of the object intensity probability density function are proposed by using the anisotropic skew normal mixture model. The proposed model not only has the capability to fit the observed data obeying a non-symmetric distribution, but also can reduce the effect of noise while preserving the objects details. Finally, expectation maximization (EM) algorithm is adopted to estimate the model parameters in order to maximize the log-likelihood function. The experiment results on synthetic images and natural grayscale images demonstrate the superior performance of the proposed model compared with other state-of-the-art segmentation methods.

A competitive chain-based Harris Hawks Optimizer for global optimization and multi-level image thresholding problems

This paper presents an enhanced Harris Hawks Optimizer (HHO) to tackle global optimization and determine the optimal threshold values for multi-level image segmentation problems. HHO is a new swarm-based metaheuristic technique that simulates the behaviors of Harris hawks during the process of catching the rabbits. The HHO established its strong performance as a swarm-based optimization technique. However, population-based HHO still may face some limitations in dealing with more multi-modal and composition problems. For example, this optimizer may be stagnated to local optima and turned to immature convergence when performing phases of exploration and exploitation. To mitigate these drawbacks, an improved HHO is proposed that considers the salp swarm algorithm (SSA) as a competitive method to enhance the balance between its exploration and exploitation trends. Firstly, a set of solutions is generated. Then, we divide those solutions into two halves, where the exploratory and exploitative phases of HHO will be applied to the first half, and the searching stages of SSA will be used to update the solutions in the second half. Thereafter, the best solutions from the union sub-populations are selected to continue the iterative process. According to the improved HHO, which is called HHOSSA, an effective multi-level image segmentation approach is also developed in this research. A comprehensive set of experiments are performed using 36 IEEE CEC 2005 benchmark functions and 11 natural gray-scale images. Extensive results and comparisons show the high ability of the SSA to improve the HHO’s performance since the proposed HHOSSA achieves a more stable performance compared to HHO, SSA, and many other well-known methods.

An Improved Marine Predators Algorithm With Fuzzy Entropy for Multi-Level Thresholding: Real World Example of COVID-19 CT Image Segmentation

Medical imaging techniques play a critical role in diagnosing diseases and patient healthcare. They help in treatment, diagnosis, and early detection. Image segmentation is one of the most important steps in processing medical images, and it has been widely used in many applications. Multi-level thresholding (MLT) is considered as one of the simplest and most effective image segmentation techniques. Traditional approaches apply histogram methods; however, these methods face some challenges. In recent years, swarm intelligence methods have been leveraged in MLT, which is considered an NP-hard problem. One of the main drawbacks of the SI methods is when searching for optimum solutions, and some may get stuck in local optima. This because during the run of SI methods, they create random sequences among different operators. In this study, we propose a hybrid SI based approach that combines the features of two SI methods, marine predators algorithm (MPA) and moth-?ame optimization (MFO). The proposed approach is called MPAMFO, in which, the MFO is utilized as a local search method for MPA to avoid trapping at local optima. The MPAMFO is proposed as an MLT approach for image segmentation, which showed excellent performance in all experiments. To test the performance of MPAMFO, two experiments were carried out. The first one is to segment ten natural gray-scale images. The second experiment tested the MPAMFO for a real-world application, such as CT images of COVID-19. Therefore, thirteen CT images were used to test the performance of MPAMFO. Furthermore, extensive comparisons with several SI methods have been implemented to examine the quality and the performance of the MPAMFO. Overall experimental results confirm that the MPAMFO is an efficient MLT approach that approved its superiority over other existing methods.

Multi-level thresholding-based grey scale image segmentation using multi-objective multi-verse optimizer

Image segmentation is among the most important techniques in image processing, and many methods have been developed to perform this task. This paper presents a new multi-objective metaheuristic based on a multi-verse optimization algorithm to segment grayscale images via multi-level thresholding. The proposed approach involves finding an approximate Pareto-optimal set by maximizing the Kapur and Otsu objective functions. Both Kapur’s and Otsu’s methods are highly used for image segmentation performed by means of bi-level and multi-level thresholding. However, each of them has certain characteristics and limitations. Several metaheuristic approaches have been proposed in the literature to separately optimize these objective functions in terms of accuracy, whereas only a few multi-objective approaches have explored the benefits of the joint use of Kapur and Otsu’s methods. However, the computational cost of Kapur and Otsu is high and their accuracy needs to be improved. The proposed method, called Multi-objective Multi-verse Optimization, avoids these limitations. It was tested using 11 natural grayscale images and its performance was compared against three of well-known multi-objective algorithms. The results were analyzed based on two sets of measures, one to assess the performance of the proposed method as a multi-objective algorithm, and the other to evaluate the accuracy of the segmented images. The results showed that the proposed method provides a better approximation to the optimal Pareto Front than the other algorithms in terms of hypervolume and spacing. Moreover, the quality of its segmented image is better than those of the other methods in terms of uniformity measures.

A note on patch-based low-rank minimization for fast image denoising

Patch-based low-rank minimization for image processing attracts much attention in recent years. The minimization of the matrix rank coupled with the Frobenius norm data fidelity can be solved by the hard thresholding filter with principle component analysis (PCA) or singular value decomposition (SVD). Based on this idea, we propose a patch-based low-rank minimization method for image denoising. The main denoising process is stated in three equivalent way: PCA, SVD and low-rank minimization. Compared to recent patch-based sparse representation methods, experiments demonstrate that the proposed method is rather rapid, and it is effective for a variety of natural grayscale images and color images, especially for texture parts in images. Further improvements of this method are also given. In addition, due to the simplicity of this method, we could provide an explanation of the choice of the threshold parameter, estimation of PSNR values, and give other insights into this method.

Robust non-local median filter

This paper describes a novel image filter with superior performance on detail-preserving removal of random-valued impulse noise superimposed on natural gray-scale images. The non-local means filter is in the limelight as a way of Gaussian noise removal with superior performance on detail preservation. By referring the fundamental concept of the non-local means, we had proposed a non-local median filter as a specialized way for random-valued impulse noise removal so far. In the non-local processing, the output of a filter is calculated from pixels in blocks which are similar to the block centered at a pixel of interest. As a result, aggressive noise removal is conducted without destroying the detailed structures in an original image. However, the performance of non-local processing decreases enormously in the case of high noise occurrence probability. A cause of this problem is that the superimposed noise disturbs accurate calculation of the similarity between the blocks. To cope with this problem, we propose an improved non-local median filter which is robust to the high level of corruption by introducing a new similarity measure considering possibility of being the original signal. The effectiveness and validity of the proposed method are verified in a series of experiments using natural gray-scale images.

First-order derivative-based super-resolution

A single fast super-resolution method based on first-order derivatives from neighbor pixels is proposed. The basic idea of the proposed method is to exploit a first-order derivatives component of six edge directions around a missing pixel, followed by back projection to reduce noise estimated by the difference between simulated and observed images. Using first-order derivatives as a feature, the proposed method is expected to have low computational complexity, and it can theoretically reduce blur, blocking, and ringing artifacts in edge areas compared to previous methods. Experiments were conducted using 900 natural grayscale images from the USC-SIPI Database. We evaluated the proposed and previous methods using peak signal-to-noise ratio, structural similarity, feature similarity, and computation time. Experimental results indicate that the proposed method clearly outperforms other state-of-the-art algorithms such as fast curvature-based interpolation.

Area summation is related to efficient neural representation

In the primary visual cortex (V1), increasing stimulus size first increases and then decreases the neural firing rate before reaching an asymptote [1,2]. This typical response curve, area summation function (ASF), defines canonical contextual modulation in typical V1 neuron. Previous studies suggest that contextual modulation associates with efficiency of the neural network [3,4]. Here, we studied the relationship between ASF and efficiency of system level neural activation patterns. Cavanaugh et al. [1] provided quantitative data from macaque cortex, as well as a mathematical description for the ASF. Therefore, we created a theoretical expected cortical response for an arbitrary gray-level image input stimulus based on ASF. Next, we used a biophysically meaningful (biomimetic) network of exponential integrate-and-fire neurons to stimulate V1 response (Heikkinen, Sharifian, Vigario and Vanni, unpublished observations). As an outcome, we compared the distance between the biomimetic simulator outputs to the expected ASF function (Distance to Area Summation, DAS). Finally, we studied entropy per spike, energy consumption and neural population sparseness as a function of the DAS for 20 natural grayscale images. In the biomimetic simulation we left V1-extrastriate and V1 excitatory-inhibitory connection strengths as free parameters in the simulations. A square search with 626 combinations of values was run for all the 20 natural stimuli resulting in 20 * 626 = 12520 simulated response patterns. Our results show that: i) there is a clear association between the entropy per spike and DAS (distance correlation = 0.66, Fig 1.a). ii) A negative link exists between the level of energy consumption (linearly associated with spike frequency) in modeled output patterns and DAS (r = 0.62, p < 0.001, Fig 1.b). iii) Population sparseness of the modeled output patterns is positively related to the DAS (r = 0.60, p < 0.001, Fig 1.c). Figure 1 In conclusion, the results of our study suggest that there is a link between area summation function and efficient information coding in the cortical visual system.

Switching non-local median filter

This paper describes a novel image filtering method for removal of random-valued impulse noise superimposed on grayscale images. Generally, it is well known that switching-type median filters are effective for impulse noise removal. In this paper, we propose a more sophisticated switching-type impulse noise removal method in terms of detail-preserving performance. Specifically, the noise detector of the proposed method finds out noise-corrupted pixels by focusing attention on the difference between the value of a pixel of interest (POI) and the median of its neighboring pixel values, and on the POI’s isolation tendency from the surrounding pixels. Furthermore, the removal of the detected noise is performed by the newly proposed median filter based on non-local processing, which has superior detail-preservation capability compared to the conventional median filter. The effectiveness and the validity of the proposed method are verified by some experiments using natural grayscale images.

A New QIM-Based Watermarking Method Robust to Gain Attack

We propose a new watermarking method based on quantization index modulation. A concept of initial data loss is introduced in order to increase capacity of the watermarking channel under high intensity additive white Gaussian noise. According to the concept some samples in predefined positions are ignored even though this produces errors in the initial stage of watermark embedding. The proposed method also exploits a new form of distribution of quantized samples where samples that interpret “0” and “1” have differently shaped probability density functions. Compared to well-known watermarking schemes, this provides an increase of capacity under noise attack and introduces a distinctive feature. Two criteria are proposed that express the feature numerically. The criteria are utilized by a procedure for estimation of a gain factor after possible gain attack. Several state-of-the-art quantization-based watermarking methods were used for comparison on a set of natural grayscale images. The superiority of the proposed method has been confirmed for different types of popular attacks.

Advances in Hyperspectral Image Classification: Earth Monitoring with Statistical Learning Methods

Hyperspectral images show similar statistical properties to natural grayscale or color photographic images. However, the classification of hyperspectral images is more challenging because of the very high dimensionality of the pixels and the small number of labeled examples typically available for learning. These peculiarities lead to particular signal processing problems, mainly characterized by indetermination and complex manifolds. The framework of statistical learning has gained popularity in the last decade. New methods have been presented to account for the spatial homogeneity of images, to include user's interaction via active learning, to take advantage of the manifold structure with semisupervised learning, to extract and encode invariances, or to adapt classifiers and image representations to unseen yet similar scenes. This tutuorial reviews the main advances for hyperspectral remote sensing image classification through illustrative examples.

데이터 은닉과 멀티플렉서 기법을 이용한 (2, 2) 비밀 공유방법

We presents a novel (2, 2) secret sharing (SS) scheme for all grayscale images. Generally, a secret image is distribute more than two shadow images, which are dealt out among participants. In order to find out secret image, participants print shadow images to transparent papers. Then, a secret image will appear as stacking transparent papers. The secret sharing scheme in this paper distribute secret image into natural grayscale images using multiplexer and data hiding scheme. After then, two participant have two shadow images respectively. The merit of the proposed scheme is that shadow images have small loss in aspect of the quality with steganographic features. Therefore, the proposed secret sharing scheme in this paper is not easily detected by attackers. The experiment result verified that the proposed scheme, obviously outperforms previous SS schemes.

Robust t-distribution mixture modeling via spatially directional information

Finite mixture model (FMM) has been successfully applied to many practical applications in recent years. However, a significant shortcoming of the FMM with Gaussian distribution is that it is sensitive to noise. Recently, Student’s t-distribution with a heavier-tailed acting as a robust alternative to Gaussian distribution is getting more and more attentions. In this paper, we propose a new Student’s t-distribution finite mixture model which incorporates the spatial relationships between the pixels and simultaneously imposes spatial smoothness. In addition, the pixel’s neighbor directional information is also integrated into the proposed model. Furthermore, the pixels’ label probability proportions are explicitly represented as probability vectors to reduce the computational costs of the proposed model. We use the gradient descend method to estimate the unknown parameters of the proposed model. Comprehensive experiments are conducted on both synthetic and natural grayscale images. The experimental results demonstrate the superiority of the proposed model over some existing models.

An Extension of the Student's t-distribution Mixture Model and the Gradient Descent method

A new Student's t-distribution finite mixture model is proposed which incorporates the local spatial information of the pixels. The pixels' label probability proportions are explicitly modelled as probability vectors in the proposed model. We use the gradient descend method to estimate the parameters of the proposed model. Compre- hensive experiments are performed for synthetic and natural grayscale images. The experimental results demonstrate that the superiority of the proposed model over some other models. Index Terms—Spatially variant finite mixture model, Stu- dent's t-distribution, Image segmentation, Gradient descent

EFFECTS OF BIT DEPTH ON THE MULTIFRACTAL ANALYSIS OF GRAYSCALE IMAGES

Multifractal box counting analysis has been widely applied to study the scaling characteristics of grayscale images. Since bit depth is an important property of such images it is desirable know the impact of varying bit depths on the estimation of the generalized dimensions (Dq). We generated random geometrical multifractal grayscale fields, which were then transformed from double precision to 16, 13, and 8 bit depths. Digitized grayscale images of soil thin sections at 13 bit depth were also selected for study and transformed to 8 bit depth. The moment based box counting method was applied to evaluate the bit depth effects on Dq. The partition functions for the multifractal fields became noticeably nonlinear on a log-log scale when q ≪ 0 as the bit depth decreased. This trend can be attributed to loss of grayscale details, changes in the local mass distribution, and the occurrence of zeros due to the bit depth transformation and data normalization processes. These effects were most pronounced for positively skewed multifractal fields, with a high proportion of extremely small mass fractions. As a result, the generalized dimensions estimated by linear regression were not always accurate, and an alternative method based on numerical derivatives was explored. The numerical method significantly improved the accuracy of the multifractal analyses; the maximum absolute difference between the analytical and numerically-derived estimates of Dq was only 9.62 × 10-3. However, when applied to situations in which the box counting scale factor did not match the scale factor used to generate the multifractal field, the numerically-derived estimates of Dq were severely biased. In this case, the linear regression method is preferable even though some error may occur due to limited bit depths. All of the soil grayscale images exhibited multifractal scaling characteristics, although there was little effect of bit depth on the resulting Dq values. Because of random fluctuations in the partition functions, the linear regression method proved to be more robust than the numerical derivative method for estimating the generalized dimensions of natural grayscale images.