Group Of Pixels Research Articles

Exploiting Laguerre transform in image steganography

Steganography, an approach used to conceal information into the digital media, generally works in two domains: spatial and transform. Though spatial domain methods are simpler, but transform domain methods are good at identifying the features which make the end system more secure. In this work, a novel Steganographic scheme based on an integer sequence named Laguerre transform (LT) is proposed. The Cover image is decomposed into non-overlapping m-pixel groups and then each such pixel group is transformed by applying LT. Variable length bits from the secret information are fabricated into the transformed components. A post-embedding adjustment is applied over these components to minimize the distortion. By applying Inverse LT (ILT), the m-pixel groups are re-computed from the resulting adjusted components. Experimental results reveal that disparity between cover and stego-pixels increases as m increases. Proposed scheme offers better stego image and higher payload compared to some state-of-the-art techniques. Code of this method is publicly available here.

A Large-Scale Multimodal CMOS Biosensor Array With 131,072 Pixels and Code-Division Multiplexed Readout

This letter presents a large-scale fully integrated multimodal sensor array for biological imaging. The 512×256 sensor array can perform spatially resolved electrochemical impedance spectroscopy (EIS) with switching frequencies up to 100 MHz, acquire multicolor optical images, and sense pH using titanium nitride (TiN) ion sensitive field effect transistors (ISFETs). The chip features code-division multiplexed (CDM) readout of groups of pixels simultaneously, enabling extended integration times at a given frame rate. The system is implemented in 180-nm CMOS with 9.5 μm × 11.5 μm pixels. Its overall fill factor is 57%, including peripheral control and readout circuits, yielding a widefield spatially resolved multimodal biosensing platform for advanced cell culture applications.

Graph-based segmentation with homogeneous hue and texture vertices

This work presents an automated segmentation method, based on graph theory, which processes superpixels that exhibit spatially similarities in hue and texture pixel groups, rather than individual pixels. The graph shortest path includes a chain of neighboring superpixels which have minimal intensity changes. This method reduces graphics computational complexity because it provides large decreases in the number of vertices as the superpixel size increases. For the starting vertex prediction, the boundary pixel in first column which is included in this starting vertex is predicted by a trained deep neural network formulated as a regression task. By formulating the problem as a regression scheme, the computational burden is decreased in comparison with classifying each pixel in the entire image. This feasibility approach, when applied as a preliminary study in electron microscopy and optical coherence tomography images, demonstrated high measures of accuracy: 0.9670 for the electron microscopy image and 0.9930 for vitreous/nerve-fiber and inner-segment/outer-segment layer segmentations in the optical coherence tomography image.

AdaBoosted Extra Trees Classifier for Object-Based Multispectral Image Classification of Urban Fringe Area

In the past decade, it is proven that satellite image classification using an object-based technique is better than the standard pixel-based technique. With the increasing need for classifying multispectral satellite images for urban planning, the accuracy of the classification becomes a significant performance parameter. Object-based classification (OBC) is a technique in which group of pixels having similar spectral properties, called objects, are generated using image segmentation and then these objects are classified based on their attributes. In this paper, the combination of a multiclass AdaBoost algorithm with extra trees classifier (ETC) is proposed with higher prediction accuracy for the OBC of the urban fringe area. The performance of the AdaBoost algorithm is found to be better in terms of classification accuracy than benchmarked SVM and RF classifiers for OBC. These classification methods were applied to IRS-R2 LISS IV data. The AdaBoosted extra trees classifier (ABETC) has demonstrated the highest accuracy with overall accuracy (OA) of 88.47% and a kappa coefficient of 0.85. The computational time of the ABETC is found to be much smaller than the RF algorithm. In detail, the sensitivity of the classifiers was investigated using stratified random sampling with various sample sizes.

A Novel DIBR 3D Image Hashing Scheme Based on Pixel Grouping and NMF

Most of the traditional 2D image hashing schemes do not take into account the change of viewpoint when constructing the final hash vector. This result in the classification accuracy rate is unsatisfactory when applied for depth-image-based rendering (DBIR) 3D image identification. In this work, pixel grouping based on histogram shape and nonnegative matrix factorization (NMF) are applied to design DIBR 3D image hashing with better robustness resisting to geometric distortions and higher classification accuracy rate for virtual image identification. Experiments show that the proposed hashing is robust against common signal and geometric distortion attacks, such as additive noise, blurring, JPEG compression, scaling, and rotation. Compared with the state-of-art schemes of traditional 2D image hashing, the proposed hashing achieves better performances under above attacks, especially for virtual image identification.

DEVELOPMENT OF SOFTWARE FOR ANALYSIS OF BEHAVIOR AND WELFARE OF BROILERS

ABSTRACT The objective of this study was to develop a software based on image processing and computer vision techniques for monitoring the feeding/collective behavior of broilers (Cobb) and validate it based on the results obtained from the visual analysis of an expert. The visual analysis was performed based on the observation and quantification of behaviors in the interval of 10 min at each hour of the day, in the period of 24 hours, totaling 1728 frames/day, for males and females. The software was developed using the Hoshen-Kopelman algorithm, for labeling clusters, which would basically be the grouping of similar pixels. This software is written in the 1999 standard of the C language. After this stage of programming, the Hoshen-Kopelman algorithm identified the birds in the region of interest (feeder and drinker), with removal of the background color to obtain the descriptors, using the same range of visual analysis. Then, the software was validated through linear regression analysis, using the R platform. Based on the correlation analyses, it was found that the coefficient of determination (R2) ranged from 0.74 to 0.97 for all software validation events. The number of broilers in the regions of interest showed R2 above 0.70 for females and above 0.89 for males, which allowed the characterization of the ingestive behavior of broilers by computer vision. Some factors that caused interference in the accuracy of the software were decisive for the result, especially the arrangement of the cameras, uneven lighting, obstruction caused by the lighting system, pendular movement of the feeders and drinker, and shade generated by them. Despite the highlighted interferences, the results allowed us to infer that the software identified and quantified the feeding behavior of broilers in an appropriate manner.

Pixel Grouping of Digital Images for Reversible Data Hiding

Pixel Grouping (PG) of digital images has been a key consideration in recent development of the Reversible Data Hiding (RDH) schemes. While a PG kernel with neighborhood pixels helps compute image groups for better embedding rate-distortion performance, only horizontal neighborhood pixel group of size 1×3 has so far been considered. In this paper, we formulate PG kernels of sizes 3×1, 2×3 and 3×2 and investigate their effect on the rate-distortion performance of a prominent PG-based RDH scheme. Specially, a kernel of size 3×2 (or 2×3) that creates a pair of pixel-trios having triangular shape and offers a greater possible correlation among the pixels. This kernel thus can be better utilized for improving a PG-based RDH scheme. Considering this, we develop and present an improved PG-based RDH scheme and the computational models of its key processes. Experimental results demonstrated that our proposed RDH scheme offers reasonably better embedding rate-distortion performance than the original scheme.

Highly sensitive SWIR detector array based on nanoscale phototransistors integrated on CMOS readout

Ultra-sensitive and fast infrared imaging has become increasingly important in applications that require high frame rates at low light levels, such as exoplanet imaging. The sensitivity of conventional short-wave infrared cameras is limited by their readout noise level. This limitation can be addressed by the internal gain of the sensors, but only if fast response time and low dark current are achieved simultaneously. Recent theoretical predictions suggested that reducing the internal capacitance of detectors with internal gain can increase their sensitivity. Here, we show the experimental validation of this prediction for III–V heterojunction phototransistors. We have fabricated a 320 × 256 array of InGaAs/InP infrared phototransistors integrated with a conventional silicon readout circuit. The array is made of two groups of pixels: 50% are devices with a 1 μm base diameter and the other 50% with a 2 μm base diameter. Characterization of a large number of pixels shows that 1 μm devices have significantly higher sensitivity than 2 μm devices. These have an average noise equivalent photon sensitivity of about 20 photons at a camera frame rate of ∼500 frames per second, which is better than the best existing infrared cameras with a similar cutoff wavelength and frame rate. Interestingly, the processing variation in the 1 μm devices resulted in variation in sensitivity, and a good number of devices show sensitivity to less than 10 photons. These results suggest that the proposed phototransistors are promising for ultra-sensitive short-wavelength infrared cameras.

RPE layer detection and baseline estimation using statistical methods and randomization for classification of AMD from retinal OCT

Background and Objective: Age-related macular degeneration (AMD) is a condition of the eye that affects the aged people. Optical coherence tomography (OCT) is a diagnostic tool capable of analyzing and identifying the disease affected retinal layers with high resolution. The objective of this work is to extract the retinal pigment epithelium (RPE) layer and the baseline (natural eye curvature, particular to every patient) from retinal spectral-domain OCT (SD-OCT) images. It uses them to find the height of drusen (abnormalities) in the RPE layer and classify it as AMD or normal.Methods: In the proposed work, the contrast enhancement based adaptive denoising technique is used for speckle elimination. Pixel grouping and iterative elimination based on the knowledge of typical layer intensities and positions are used to obtain the RPE layer. Using this estimate, randomization techniques are employed, followed by polynomial fitting and drusen removal to arrive at a baseline estimate. The classification is based on the drusen height obtained by taking the difference between the RPE and baseline levels. We have used a patient, wise classification approach where a patient is classified diseased if more than a threshold number of patient images have drusen of more than a certain height. Since all slices of an affected patient will not show drusen, we are justified in adopting this technique.Results: The proposed method is tested on a public data set of 2130 images/slices, which belonged to 30 patient volumes (15 AMD and 15 Normal) and achieved an overall accuracy of 96.66%, with no false positives. In comparison with existing works, the proposed method achieved higher overall accuracy and a better baseline estimate.Conclusions: The proposed work focuses on AMD/normal classification using a statistical approach. It does not require any training. The proposed method modifies the motion restoration paradigm to obtain an application-specific denoising algorithm. The existing RPE detection algorithm is modified significantly to make it robust and applicable even to images where the RPE is not very evident/there is a significant amount of perforations (drusen). The baseline estimation algorithm employs a powerful combination of randomization, iterative polynomial fitting, and pixel elimination in contrast to mere fitting techniques. The main highlight of this work is, it achieved an exact estimation of the baseline in the retinal image compared to the existing methods.

PARALLEL K-MEANS CLUSTERING WITH NAÏVE SHARDING FOR UNSUPERVISED IMAGE SEGMENTATION VIA MPI

In digital image processing, image segmentation is an essential step in which an image is partitioned into groups of pixels. k-means clustering algorithm, which is often considered as fast and efficient, is one of the most widely used clustering algorithms to segment an image. However, as the problem size gets larger, the k-means starts to spend a significant amount of time to process. At this point, parallelization techniques should be applied to reduce the required time. Designing an efficient parallel and distributed model is not a trivial job since it should correspond to the parallel computer architecture and take communication and load balancing among processors into account. In this study, we propose a parallel and distributed k-means clustering algorithm with naive sharding centroid initialization for image segmentation. The proposed algorithm adopts the Message Passing Interface (MPI) standard to take advantage of the computational power of distributed computing nodes in a High Performance Computing Cluster. We demonstrate the parallel scalability of the proposed algorithm using up to 128 cores that achieves approximately 104 times faster clustering time.

Quantized ternary pattern and singular value decomposition for the efficient mining of sequences in SRSI images

The growth and development of particular region over time can be witnessed by remote sensing images. Although such raw images have less possibility to derive the insights, Serial Remote Sensing Images (SRSI) has the large potential to discover the patterns. The evolution of spatial patterns in various areas including urban development, expansion of vegetation cover and agriculture is the evidence for the utilization of SRSI accumulation. The application of conventional sequential pattern-mining algorithms on the SRSI images results in high computational complexity. This issue can be resolved by grouping the pixels and mining sequence patterns. A one-pass framework is introduced to compress and hide the data in the marked stream without any loss. In this paper, we proposed a Quantized ternary pattern based pixel grouping and Singular Value Decomposition—Run Length Coding based pattern mining. The algorithms are experimented using a dataset, namely, the Cropland data layer dataset. The proposed algorithm is efficient in terms of mining time and sequence pattern generation.

Automatic Image Pixel Clustering based on Mussels Wandering Optimization

Image pixel clustering or segmentation intends to identify pixel groups on an image without any preliminary labels. It remains a challenging task in computer vision since the size and shape of object segments are varied. Moreover, determining the segment number in an image without prior knowledge of the image content is an NP-hard problem. In this paper, we present an automatic image pixel clustering scheme based on mussels wandering optimization. An activation variable is applied to determine the number of clusters automatically with the cluster centers optimization. We revise the within- and between-class sum of squares ratio for random natural image content and develop a novel fitness function for the image pixel clustering task. Our proposed scheme is compared against existing state-of-the-art techniques using both synthetic data and real ASD dataset. Experimental results show the superiority performance of the proposed scheme.

Land change syndromes identification in temperate forests of Hindukush Himalaya Karakorum (HHK) mountain ranges

ABSTRACT In this study, we analyzed time series of a vegetation index to identify land-cover/land-use changes in HHK mountain regions administered by Pakistan. Monthly MODIS (Moderate Resolution Imaging Spectroradiometer) EVI (Enhanced Vegetation Index) series were decomposed to retrieve changes in long-term mean, peaking magnitude and seasonal characteristics. Resulting linear trend patterns were used as a baseline to map syndromes over the study area. To address non-linear changes, inter-annual variation (IAV) and short-term variation (STV) was also computed. Distributed lag-models (DLM) were used to determine significant correlation between GIMMS (Global Inventory Monitoring and Mapping) NDVI (Normalized Difference Vegetation Index) and rainfall as an additional syndrome to highlight climate sensitive regions. Our results indicate that land use is mainly controlled by two factors: elevation and river network. Based on that, there is a particular spatial distribution of agricultural intensification because of deforestation, forest degradation from unsustainable use, forest regeneration and human settlements near rivers. Most of the trends observed showed a persistent greening pattern compared to small groups of pixels with negative trends. Outcomes of DLM do not provide plausible links between rain and forest biomass. It does however suggest a positive response of crops to rainfall events in the arid zones of the study area. High short-term fluctuation in EVI residuals occurred in areas that experienced considerable land modification and where land-use patterns are not stable. IAV was high in regions around rivers and water works and its impact on forest dynamics could not be substantiated. Land change patterns described here can be used by decision makers for forest restoration programmes.

Spectral–Spatial Joint Sparse NMF for Hyperspectral Unmixing

The nonnegative matrix factorization (NMF) combining with spatial–spectral contextual information is an important technique for extracting endmembers and abundances of hyperspectral image (HSI). Most methods constrain unmixing by the local spatial position relationship of pixels or search spectral correlation globally by treating pixels as an independent point in HSI. Unfortunately, they ignore the complex distribution of substance and rich contextual information, which makes them effective in limited cases. In this article, we propose a novel unmixing method via two types of self-similarity to constrain sparse NMF. First, we explore the spatial similarity patch structure of data on the whole image to construct the spatial global self-similarity group between pixels. And according to the regional continuity of the feature distribution, the spectral local self-similarity group of pixels is created inside the superpixel. Then based on the sparse expression of the pixel in the subspace, we sparsely encode the pixels in the same spatial group and spectral group respectively. Finally, the abundance of pixels within each group is forced to be similar to constrain the NMF unmixing framework. Experiments on synthetic and real data fully demonstrate the superiority of our method over other existing methods.

AN ADAPTIVE IMAGE SCALING ALGORITHM BASED ON CONTINUOUS FRACTION INTERPOLATION AND MULTI-RESOLUTION HIERARCHY PROCESSING

Traditional interpolation algorithms often blur the edges of the target image due to low-pass filtering effects, making it difficult to obtain satisfactory visual effects. Especially when the reduction ratio becomes small, the phenomenon of jagged edges and partial information loss will occur. In order to obtain better image scaling quality, an adaptive image scaling algorithm based on continuous fraction interpolation and multi-resolution hierarchical processing is proposed. In order to overcome the noise problem of the original image, this paper first performs wavelet decomposition on the original image to obtain multiple images with different resolutions. Secondly, in order to eliminate the influence of local area variance on the overall image, weighted average is performed on images of different resolutions. Then, in order to overcome the blurring effect of the weighted average image, by calculating the variance of the three groups of pixels around the target pixel, selecting a group of pixels with the smallest variance and using the Salzer continuous fraction interpolation equation to obtain the gray value of the target pixel. Finally, the multiple corrected images are stitched together into a scaled image. The algorithm in this paper achieves a high-order smooth transition between pixels in the same feature area, and can also adaptively modify the pixels of the image. The experimental results show that the edge of the target image obtained by the algorithm in this paper is clear, and the algorithm complexity is low, which is convenient for hardware implementation and can realize real-time image scaling.

An Algorithm of Image Segmentation Based on Persistent Homology for Solving Defects Searching Problems

Purpose of research is to develop an image segmentation algorithm based on the persistent homology for solving problems of searching and classifying defects. The algorithm is aimed at improving the quality of products at enterprises with continuous production (metallurgy, woodworking, and others).Methods. To segment an image, it is proposed to specify links between pixels in the image. In the future, during the iterative breaking of links, as their weights increase, pixels will be combined into groups called holes. Pixels that are in a single group have both their original characteristics and characteristics common for the entire group, and they also change the weights of their links with representatives of other groups. This creates a history of the formation of separate groups of pixels which can be specified as segments with a time-based characteristic of the change.Results. The result of the research is the development of an algorithm designed to search for and classify defects in various materials. The optimal algorithm for applying the principle of persistent homology to images has been developed, and factors determining the transition boundaries of image objects have been analyzed and selected. The segmentation algorithm was tested on metal images obtained from sheet rolling equipment. The results of comparing the proposed algorithm with the K-means and Mean-Shift segmentation algorithms for different parameters are provided in the article.Conclusion. Using persistent homology in image segmentation tasks can enable creating a tool that can be applied to materials with different structures without any need for significant changes. The software implementation of the segmentation process based on the principles of computer topology has shown high flexibility due to the storing of the history of segment changes.

Noise suppressed and bias field corrected image segmentation method for porous Ni-YSZ anode microstructure

Accurate three-phase identification from Solid Oxide Fuel Cell (SOFC) anode micrograph is challenged by both noise and intensity inhomogeneity. In this paper, a novel framework is proposed for porous Ni-YSZ cermet anode Optical Microscopy (OM) image segmentation. The proposed framework takes advantage of a statistical model in which an observed image is decomposed into two multiplicative components (bias field and true image) and one additive component (noise). A two-stage Principal Component Analysis with Local Pixel Grouping (LPG-PCA) denoising algorithm is firstly performed to suppress additive noise, it can preserve more image local structural features by modeling a pixel and its nearest neighbors as a vector variable and selecting training samples with similar contents to this variable in a local window for PCA transformation. In order to enhance the robustness to noise, uneven illumination and other outliers, a kernel metric is introduced into fuzzy clustering method embedded with bias field correction for image segmentation. The proposed method has been compared to other state-of-the-art segmentation algorithms on both simulated images and real SOFC anode OM images. Extensive experiments results have demonstrated that the proposed framework can successfully eliminate the influence of both uneven illumination and noise on real Ni/YSZ anode OM images to obtain a better three-phase identification accuracy. The high-quality segmentation results lay firm foundation for accurate microstructural parameter extraction.

Development of Deep Learning Algorithm using Convolutional Neural Network for Medical Imaging

Medical imaging is the procedure and approach of formulating graphic models of the peculiarity of a body system for medical investigation and treatment, and also graphical illustration of the function of several internal organs or structures. To identify the affected tissues of the brain in a case of brain tumors, it is important to get high precision and accuracy to locate exact pixels. Manual analysis may be erroneous and so it is important to use deep learning image segmentation technique. Segmentation of graphic is the technique of dividing a graphic in to several group of pixels. The earlier objective of the segmentation is actually to produce details much easier and enhance the manifestation of clinical images into significant content. Segmentation is a complicated activity due to the excessive variability in the graphics. The computational intelligence is modern way for application automation. Existing studies shows need of deep learning research for fast and accurate medical imaging solutions. Hence, this paper presents the CNN framework (for an analysis of brain tumors) as a base for further research methodology development. The paper also provides a pilot research analysis that can further be used to develop improved precision and visibility.

Gradient-driven unsupervised video segmentation using deep learning techniques

We propose a three-dimensional video segmentation method using deep learning convolutional neural nets. The algorithm utilizes the local gradient computed at each pixel location together with the global boundary map acquired through deep learning methods to generate initial pixel groups by traversing from low to high gradient regions. A local clustering method is then employed to refine these initial pixel groups. The refined subvolumes in the homogeneous regions of video are selected as initial seeds and iteratively combined with adjacent groups based on intensity similarities. The volume growth is terminated at the color boundaries of the video. The oversegments obtained from the above steps are then merged hierarchically by a multivariate approach yielding a final segmentation map for each frame. The results show that our proposed methodology compares favorably well, on a qualitative and quantitative level, in segmentation quality and computational efficiency, with the latest state-of-the-art techniques utilizing the video segmentation benchmark dataset.

A thumbnail-based hierarchical fuzzy clustering algorithm for SAR image segmentation

This paper proposes a novel algorithm for segmentation of synthetic aperture radar (SAR) image, our proposed algorithm (THFCM) is based on thumbnail representations and a hierarchical fuzzy C-means (FCM) approach. THFCM firstly divides the image into pixel groups to extract local feature, and the major pixels of each pixel group are selected to construct a thumbnail. FCM is then used to segment each thumbnail, and hierarchical segmentation is then performed on the overall image data, based on the results of thumbnail clustering. The thumbnail approach leverages local image information, helping to overcome speckle noise, while the hierarchical approach improves computational efficiency. Experiments on simulated and real SAR images suggest that THFCM outperforms several other state-of-the-art algorithms in terms of both segmentation accuracy and running time.