Original Pixel Research Articles

Image Warp Preserving Content Intensity

An accurate method for warping images is presented. Differently from most commonly used techniques, this method guarantees the conservation of the intensity of the transformed image, evaluated as the sum of its pixel values over the whole image or over corresponding transformed subregions of it. Such property is mandatory for quantitative analysis, as, for instance, when deformed images are used to assess radiances, to measure optical fluxes from light sources, or to characterize material optical densities. The proposed method enforces area resampling by decomposing each rectangular pixel in two triangles, and projecting the pixel intensity onto half pixels of the transformed image, with weights proportional to the area of overlap of the triangular half-pixels. The result is quantitatively exact, as long as the original pixel value is assumed to represent a constant image density within the pixel area, and as long as the coordinate transformation is diffeomorphic. Implementation details and possible variations of the method are discussed.

De-noising of salt and pepper noise using deep learning-based alpha-guided grey wolf optimization

The real images are undermined by salt and pepper noise due to uproarious sensors or communication errors. In this paper, a hybrid SAR image noise reduction using the adaptive pulse-coupled neural network (APCNN) which is optimized by an alpha-guided grey wolf optimizer (AgGWO) in the shearlet transform domain has been proposed. The shearlet transform is utilized to decompose the input SAR image. After the completion of AgGWO optimization, PCNN filtering strategy has been utilized to supplant the noisy pixels into related information pixel components, from which a restoration of the noise reduced images can be obtained. This proposed methodology efficiently resolves the difficulties that emerge from the basic PCNN parameter determination problem. In this methodology, the noisy pixels are secluded well from the image and original noiseless pixels are reestablished well, which can prompt better conservation of edges of an image. This proposed APCNN-AgGWO method has also been compared with other existing noise reduction methods and it yields superior de-noising impact, in terms of structural similarity index measure (SSIM) of 98.46%, peak signal-to-noise ratio (PSNR) of 39.49%, and standard deviation (STD) of 38.64%.

Innovation and Research of Digital Artwork Design Based on Big Data and Integrated Media

We are in the era of big data, and art, which has developed along with human history, has been given new vehicles and forms of creation with the use of integrated media technology. This study first introduces artwork and its definition and classification, then briefly describes the definition and development of digital art, and focuses on the development, contribution, and drawbacks of digital artwork NFT (nonfunctional token). This study divided the design model into four layers based on the neural network calculation of factorization machine to build, and the data of each layer was analyzed and processed to build the FNN algorithm model. It was found that the largest proportion of NFT artworks circulating in the trading market is in paintings, at 55%. The HSV color space was selected as the color perception model, and then the picture pixel positions were determined, and redundant pixels were eliminated. To avoid data overflow, the spatial memory occupied by the pixels of the painting was calculated. The results showed that the growth of pixel space complexity was smooth and converged to a straight line, indicating that the original pixel data extracted in this experiment was stable. The circle was selected as the basic guideline for the creation of four digital art paintings with an overall uniform but innovative style.

Compound adversarial examples in deep neural networks

Although deep learning has made great progress in many fields, they are still vulnerable to adversarial examples. Many methods for generating adversarial examples have been proposed, which either contain adversarial perturbation or patch. In this paper, we explore the method that creates compound adversarial examples including both perturbation and patch. We show that fusing two weak attack modes can produce more powerful adversarial examples, where the patch covers only 1% of the pixels at random location in the image, and the perturbation changes only by 2/255 in the original pixel value (scale to 0–1). For both targeted attack and untargeted attack, compound attack can improve the generative efficiency of adversarial examples, and can attain higher attack success rate with fewer iteration steps. The compound adversarial examples successfully attack the models with defensive mechanisms that previously can defend perturbation attack or patch attack. Furthermore, the compound adversarial examples show good transferability on normal trained classifiers and adversarial trained classifiers. Experimental results on a series of widely used classifiers and defense models show that the proposed compound adversarial examples have strong robustness, high effectiveness, and good transferability.

Nondestructive Classification of Maize Moldy Seeds by Hyperspectral Imaging and Optimal Machine Learning Algorithms.

Mildew of maize seeds may affect their germination rates and reduce crop quality. It is crucial to classify maize seeds efficiently and without destroying their original structure. This study aimed to establish hyperspectral datasets using hyperspectral imaging (HSI) of maize seeds with different degrees of mildew and then classify them using spectral characteristics and machine learning algorithms. Initially, the images were processed with Otus and morphological operations. Each seed’s spectral features were extracted based on its coding, its edge, region of interest (ROI), and original pixel coding. Random forest (RF) models were optimized using the sparrow search algorithm (SSA), which is incapable of escaping the local optimum; hence, it was optimized using a modified reverse sparrow search algorithm (JYSSA) strategy. This reverse strategy selects the top 10% as the elite group, allowing us to escape from local optima while simultaneously expanding the range of the sparrow search algorithm’s optimal solution. Finally, the JYSSA-RF algorithm was applied to the validation set, with 96% classification accuracy, 100% precision, and a 93% recall rate. This study provides novel ideas for future nondestructive detection of seeds and moldy seed selection by combining hyperspectral imaging and JYSSA algorithms based on optimized RF.

Efficient Color Quantization Using Superpixels.

We propose three methods for the color quantization of superpixel images. Prior to the application of each method, the target image is first segmented into a finite number of superpixels by grouping the pixels that are similar in color. The color of a superpixel is given by the arithmetic mean of the colors of all constituent pixels. Following this, the superpixels are quantized using common splitting or clustering methods, such as median cut, k-means, and fuzzy c-means. In this manner, a color palette is generated while the original pixel image undergoes color mapping. The effectiveness of each proposed superpixel method is validated via experimentation using different color images. We compare the proposed methods with state-of-the-art color quantization methods. The results show significantly decreased computation time along with high quality of the quantized images. However, a multi-index evaluation process shows that the image quality is slightly worse than that obtained via pixel methods.

ImageGCN: Multi-Relational Image Graph Convolutional Networks for Disease Identification With Chest X-Rays.

Image representation is a fundamental task in computer vision. However, most of the existing approaches for image representation ignore the relations between images and consider each input image independently. Intuitively, relations between images can help to understand the images and maintain model consistency over related images, leading to better explainability. In this paper, we consider modeling the image-level relations to generate more informative image representations, and propose ImageGCN, an end-to-end graph convolutional network framework for inductive multi-relational image modeling. We apply ImageGCN to chest X-ray images where rich relational information is available for disease identification. Unlike previous image representation models, ImageGCN learns the representation of an image using both its original pixel features and its relationship with other images. Besides learning informative representations for images, ImageGCN can also be used for object detection in a weakly supervised manner. The experimental results on 3 open-source x-ray datasets, ChestX-ray14, CheXpert and MIMIC-CXR demonstrate that ImageGCN can outperform respective baselines in both disease identification and localization tasks and can achieve comparable and often better results than the state-of-the-art methods.

A New Bilinear Supervised Neighborhood Discrete Discriminant Hashing

Feature extraction is an important part of perceptual hashing. How to compress the robust features of images into hash codes has become a hot research topic. Converting a two-dimensional image into a one-dimensional descriptor requires a higher computational cost and is not optimal. In order to maintain the internal feature structure of the original two-dimensional image, a new Bilinear Supervised Neighborhood Discrete Discriminant Hashing (BNDDH) algorithm is proposed in this paper. Firstly, the algorithm constructs two new neighborhood graphs to maintain the geometric relationship between samples and reduces the quantization loss by directly constraining the hash codes. Secondly, two small rotation matrices are used to realize the bilinear projection of the two-dimensional descriptor. Finally, the experiment verifies the performance of the BNDDH algorithm under different feature types, such as image original pixels and a Convolutional Neural Network (CNN)-based AlexConv5 feature. The experimental results and discussion clearly show that the proposed BNDDH algorithm is better than the existing traditional hashing algorithm and can represent the image more efficiently in this paper.

Convex optimization for additive noise reduction in quantitative complex object wave retrieval using compressive off-axis digital holographic imaging

Abstract Image denoising is one of the important problems in the research field of computer vision, artificial intelligence, 3D vision, and image processing, where the fundamental aim is to recover the original image features from a noisy contaminated image. The camera sensor additive noise present in the holographic recording process reduces the quality of the retrieved image. Even though various techniques have been developed to minimize the noise in digital holography, the noise reduction still remains a challenging task. This article presents a compressive sensing (CS) technique to minimize the additive noise in the digital holographic reconstruction process. We demonstrate the reduction of additive noise using complex wave retrieval method as a sensing matrix in the CS model. The proposed CS method to suppress the noise during the reconstruction process is illustrated using numerical simulations. Only 50% of the pixel measurements are considered in the noisy hologram, which is far less than the original complex object pixels. The impact of additive gaussian noise in the recording plane on the reconstruction accuracy of both intensity and phase distribution is analysed. The CS method denoises and estimates the complex object information accurately. The numerical simulation results have shown that the proposed CS method has effectively minimized the noise in the reconstructed image and has greatly improved the quality of both intensity and phase information.

Image Dehazing Algorithm for Relieving Halo Effect and Color Distortion in Smooth Regions

Aiming to solve the problem that the dehazing algorithm based on the dark channel produces halo effect and color distortion in smooth region, a novel dehazing algorithm is proposed using bilateral filter guided homomorphic filtering. Firstly, a nonlinear stretch model is constructed to characterize the contrast relationship among each color component before and after dehazing. The model is proportional to the difference between the original pixels and the atmospheric light value while inversely proportional to the transmittance. Secondly, the coarse transmittance map is decomposed into an atmospheric light component and a reflection component by use of the bilateral filter guided spatial homomorphic filtering. Finally, an adaptive function is defined, taking the ratio of the foreground area as a variable. It is used to calculate a refined transmittance map by linearly weighting the atmospheric light component and the reflection component, and then employed to accomplish the dehazing with the dark channel prior model. The real foggy images in O-HAZE and RESIDE datasets is used as the test data, and the UCIQE, FADE, SSIM, NIQE, and hue restoration ability are adopted as the objective evaluation matrics. The evaluation scores of the proposed algorithm increase by 21.68% on average, compared with those of the state-of-art algorithms. The results show that the proposed algorithm is able to effectively avoid the color distortion in large smooth area, as well as the halo & fake edge effects at the edges and contours, achieves higher subjective and objective quality.

Scenario prediction for power loads using a pixel convolutional neural network and an optimization strategy

Accurate and reliable prediction of power load is critical to ensure the economy and stability of power systems. However, deterministic point prediction can scarcely be accurate due to the fluctuating and stochastic behavior of power load series, resulting in high risks for the system operation. Scenario prediction is a widely used method to model stochastic behavior by generating a group of possible power load scenarios rather than deterministic point predictions, so that system operators can account for the uncertainty of power loads. In this paper, a new deep generative network-based method is proposed for scenario prediction of power loads, in which structure and parameters are redesigned on the original pixel convolutional neural network (PixelCNN). An optimization model is presented to search for a range of power load scenarios with similar shapes, temporal dependency, and probability distribution as the real ones. Numerical simulations on a real-world power load dataset show that the PixelCNN outperforms other generative networks for the scenario prediction of power loads.

CGRNet: Contour-guided graph reasoning network for ambiguous biomedical image segmentation

In this work, we propose to address the existing problem of biomedical image segmentation that often produces results, which fail to capture the exact contours of the target and suffer from ambiguity. Most previous techniques are suboptimal because they often simply concatenate contour information to alleviate this problem, while ignoring the correlation between regions and contours. As a matter of fact, the relationship between cross-domain features is an important clue for ambiguous pixel segmentation in biomedical images. To this end, we contribute a simple yet effective framework called Contour-Guided Graph Reasoning Network (CGRNet) for more accurate segmentation against ambiguity, which is capable of capturing the semantic relations between object regions and contours through graph reasoning. Specifically, we first perform a global graph representation of the low-level and high-level features extracted by the feature extractor, where clusters of pixels with similar features are mapped to each vertex. Further, we explicitly combine contour information as the geometric prior, which can aggregate features of contour pixels to graph vertices and focus on features along the boundaries. Then, the cross-domain features propagate information through the vertices on the graph to efficiently learn and reason about the semantic relations. Finally, the learned refinement graph features are projected back to the original pixel coordinate space for the final pixel-wise segmentation task. Extensive experiments on the three publicly available Kvasir, CVC-612, and COVID19-100 datasets show the effectiveness of our CGRNet with superior performance to existing state-of-the-art methods. Our code is publicly available at: https://github.com/DLWK/CGRNet.

Synchronously scrambled diffuse image encryption method based on a new cosine chaotic map

We present a new cosine chaotic mapping proved by chaos theory test and analysis such that the system has good cryptography properties, wide chaos range, simple structure, and good sensitivity to initial value, and the mapping can meet the needs of chaotic image encryption. Based on the cosine chaotic system, we propose a new encryption method. First, according to the cyclic characteristics of the mapping, the cyclic information wave is simulated. Second, the quasi-Doppler effect is used to synchronously scramble and diffuse the image to obfuscate the original pixel. Finally, the XOR diffusion of image pixels is carried out by information wave to further enhance the encryption effect. Simulation experiment and security analysis show that the algorithm has good security, can resist the common attack mode, and has good efficiency.

StereoARS: Quality Evaluation for Stereoscopic Image Retargeting With Binocular Inconsistency Detection

Many stereoscopic image retargeting (SIR) methods have been developed for automatically and intelligently resizing stereoscopic images and we cannot always rely on time-consuming subjective user studies to validate the performance of different SIR methods. It is therefore required to design reliable objective metrics for SIR quality evaluation. This paper extends our previous 2D aspect ratio similarity (ARS) metric to a stereo 3D version termed as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">StereoARS</i> where the key idea is to investigate into retargeting inconsistency between the original stereo correspondences. Our proposed <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">StereoARS</i> operates via two stages: monocular quality estimation and binocular inconsistency detection. In the first stage, monocular quality estimation is performed by applying a modified ARS measure on the left and right views separately to quantify the quality degradation within each monocular view. In the second stage, binocular inconsistency detection is performed in both pixel-level and grid-level to characterize the influence of binocular rivalry and stereo visual discomfort on SIR quality. In addition, we also measure to what extent the original pixel visibility relation is preserved after SIR as another binocular quality factor. Finally, these monocular and binocular quality estimates are fused to produce an overall SIR quality score. Extensive experiments have demonstrated that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">StereoARS</i> achieves better alignment with human subjective ratings than the existing metrics by a large margin.

Green Visual Sensor of Plant: An Energy-Efficient Compressive Video Sensing in the Internet of Things.

Internet of Things (IoT) realizes the real-time video monitoring of plant propagation or growth in the wild. However, the monitoring time is seriously limited by the battery capacity of the visual sensor, which poses a challenge to the long-working plant monitoring. Video coding is the most consuming component in a visual sensor, it is important to design an energy-efficient video codec in order to extend the time of monitoring plants. This article presents an energy-efficient Compressive Video Sensing (CVS) system to make the visual sensor green. We fuse a context-based allocation into CVS to improve the reconstruction quality with fewer computations. Especially, considering the practicality of CVS, we extract the contexts of video frames from compressive measurements but not from original pixels. Adapting to these contexts, more measurements are allocated to capture the complex structures but fewer to the simple structures. This adaptive allocation enables the low-complexity recovery algorithm to produce high-quality reconstructed video sequences. Experimental results show that by deploying the proposed context-based CVS system on the visual sensor, the rate-distortion performance is significantly improved when comparing it with some state-of-the-art methods, and the computational complexity is also reduced, resulting in a low energy consumption.

Embedding Data in Non-Important Gabor Ridges

Hiding information either by steganography or watermarking operation is essential for computer science. It is used as a method for sending secure and important information. It has special features that even if this information is discovered, the third part cannot reconstruct the original information in any way, not famously and commonly nor difficultly and especially by using any technique or algorithm. There are many hiding techniques; each one of them used different paths to ensure standard properties such as optimizing security level, increasing the amount of embedding data, decreasing the ability to reconstruct hidden information by any unwanted part. Here in the following suggested technique, different quantities of bits are hidden in various pixels depending on some discovered constraints by using the famous Gabor filter then divided ridges part in a selected fingerprint image for the purpose of hiding relevant information into important and non-important pixels that laying out of ridges in final images after applying Gabor filter which consider not important regions in fingerprint image for discovering important features from the image. The pixels used in hiding information belong to finger ridges in pure fingerprint images but covert to white pixels (valley) in the matrix after applying the Gabor filter and in original white pixels in both matrices by using different techniques for each type of pixels. The imaging stage constructed after the suggested technique is good and identical to the pure image for the accepted degree as shown in used fidelity metrics (PSNR). It is similar to a pure fingerprint matrix as proved in these metrics for calculating the level of goodness of proposed algorithms. Our algorithm exploits existing pixels in fingerprint image with different levels of importance and avoids pixels with high importance for protected important features from non-deterministic varying.

A Hybrid Clustering-Based Approach to Color Modification of Art Paintings for Protanopes and Deuteranopes

The normal human color vision is trichromatic, and it originates from the comparison of the rates at which photons are absorbed by three types of photoreceptor cone-cells contained in the eye's retina. The absence or malfunctioning of one or two types of cone results in color-blindness. This paper proposes an algorithmic framework to appropriately modify (i.e., recolor) art paintings for two types of color-blindness called protanopia and deuteranopia. The algorithmic framework employs four distinct steps applied in sequence. First, the image colors are clustered into a prespecified number of representative colors. The second step determines the representative colors that are confused by the protanopes or the deuteranopes. Third, an optimization problem is proposed to appropriately recolor the representative colors detected in the previous step. Finally, given the recolored representative colors, the original image pixels associated with those colors are also modified accordingly. The method is tested and evaluated in terms of quantitative, qualitative, and subjective comparison with three other recoloring algorithms. The results are promising in the sense that the proposed method outperforms the competitive algorithms, maintaining the overall aesthetic of the paintings.

Marginalized Graph Self-Representation for Unsupervised Hyperspectral Band Selection

Unsupervised band selection is an essential step in preprocessing hyperspectral images (HSIs) to select informative bands. Most existing methods exploit the spatial information from the entire HSI while ignoring the difference between diverse homogeneous regions. Moreover, traditional methods utilize the limited size of data for model training that may result in degraded generalization performance. In this article, we propose a marginalized graph self-representation (MGSR) method for unsupervised hyperspectral band selection. To explore the spatial information from diverse homogenous regions, MGSR generates the segmentations of an HSI by superpixel segmentation and records the relationships between adjacent pixels of the same segmentation in a structural graph. Meanwhile, to improve the generalization and robustness, infinite corrupted samples are obtained from the original pixels by introducing noises in spectral bands for model training. To solve the proposed formulation, we design an alternating optimization algorithm to marginalize out the corruption and search for the optimal solution. Experimental studies on HSI datasets demonstrate the effectiveness of the proposed MGSR and the superiority over the state-of-the-art methods. The source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/ZhangYongshan/MGSR</uri> .

Hyper-Embedder: Learning a Deep Embedder for Self-Supervised Hyperspectral Dimensionality Reduction

Hyperspectral imaging has attracted growing interest among the researchers from the geoscience and remote sensing fields owing to its very rich spectral information. However, the high spectral dimensionality of hyperspectral images tends to suffer from information redundancy. Manifold embedding is a mainstream strategy of nonlinear hyperspectral dimensionality reduction. The sensitivity to the noise and the inflexibility to the out-of-sample problem (i.e., new samples) are the main drawbacks of the manifold embedding based methods. To this end, we propose to learn a deep embedder in a self-supervised fashion for hyperspectral dimensionality reduction, called Hyper-Embedder. Hyper-Embedder effectively reduces the computational complexity and storage-costing compared to conventional embedding models, and improves the robustness against various noises, e.g., spectral variabilities. More significantly, Hyper-Embedder is capable of learning an explicit nonlinear mapping to make a one-to-one match between each original pixel (spectral signature) in the hyperspectral image and its dimension-reduced representation. These low-dimensional representations can be generated and given by existing and classic nonlinear manifold embedding methods. In this paper, we attempt to learn the correspondence or mapping by optimizing a deep regression network. The to-be-developed network can not only capture the local topological knowledge graph of all spectral signatures of hyperspectral data but be applicable to fast prediction and inference of samples from other hyperspectral scenes. The proposed Hyper-Embedder outperforms existing state-of-the-art hyperspectral dimensionality reduction algorithms on two commonly-used hyperspectral datasets, i.e., Indian Pines and Augsburg scenes.

Dynamic Negative Sampling Autoencoder for Hyperspectral Anomaly Detection

Hyperspectral anomaly detection (HAD) aims at detecting the anomalies without any prerequisite information, which gains lots of attention in recent years. Most of existing detectors locate the anomalies by eliminating the background. The background is usually reconstructed by utilizing only the global and local homogenous attribute, no matter the matrix decomposition-based or the deep learning-based methods. In this article, a dynamic negative sampling autoencoder is proposed for the hyperspectral anomaly detection (DNA-HAD). Some pixels are randomly selected and altered as negative samples. Both the rest original pixels and the altered negative samples are sent into the network. An adaptive adjusted loss function is designed to suppress the reconstruction error for the original pixels, and to enlarge the error for the negative samples. Meanwhile, skip connection is designed to ensure features at both shallow levels and deep levels being utilized for the reconstructing process. In this way, by importing some negative samples in the reconstructing process, the proposed DNA-HAD is not only robust to reconstruct the background but also sensitive to detect the anomalies. Experiments on six hyperspectral imagery which are captured by different sensors have demonstrated the effectiveness of the proposed method.