Pixel Difference Research Articles

Discriminative Binary Descriptor for Finger Vein Recognition

In this paper, we propose a new local discriminative feature learning method for finger vein recognition. Unlike most previous finger vein recognition methods, which use hand-crafted descriptors such as local binary pattern, local line binary pattern, and Gabor features, this paper aims to learn a feature mapping to enhance the discriminative ability of local features. To achieve this goal, we first extract multi-directional pixel difference vectors (MDPDVs) for each pixel in a training finger vein image by computing the difference between each pixel and its straight-line neighboring pixels. Second, we learn a feature mapping to project these MDPDVs into low-dimensional binary codes in a supervised manner, where: 1) the loss between the original real-valued codes and learned binary vectors is minimized; 2) the between-class variation of the local binary features is maximized; and 3) the within-class variation of the local binary features is minimized. Last, we represent each finger vein image as a histogram feature by clustering and pooling these binary codes. Experiments on SDUMLA-FV and PolyU databases verify the superior performance of the proposed method over other existing finger vein recognition methods.

Hyperthermia therapy monitoring with guidance of B-Mode ultrasound: Study on rat breast tumor tissue

Local hyperthermia treatment for cáncer therapy has shown increasing interest in conjunction to the other existence treatment as such chemotherapy, clinical surgery and radiation. B-Mode ultrasound have shown great potential for local hyperthermia, as it is nonionizing, convenient, and has relatively simple signal processing and it offers good spatial resolution.The aim of this study is to compare the most optimum B-Mode ultrasound parameters in normal and pathological tissue. The subjects were dissected and exposed to hyperthermia at 37◦C, 40◦C, 45◦C, 50◦C and 55◦C. Result shows, for mean grey scale in normal and pathological tissue, 37◦C and 40◦C was optimum temperature with correspond values of 188.08 and 199.26 meanwhile for pixel deviation analysis, at 45◦C and 55◦C the best temperature is observed in both tissue with pixel deviation of 57.27 and 57.25. Additionally, it is proven with accuracy of testing to validation data using Artificial Neural Network, ANN, about 88.89% and 81.25% was achieved.

A restoration method for distorted comics to improve comic contents identification

In recent years, copyright violations due to the illegal copying and distribution of e-comic contents have become an important issue. Although such violations can be rapidly and reliably detected by fingerprinting techniques, scanning introduces photometric and geometric distortions. This paper presents a restoration framework for reducing photometric and geometric distortions in copied comics to improve their content identification. First, the photometric distortion is reduced by conventional homomorphic filtering with histogram equalization. Next, the corner detection is improved by a consecutive pixel difference method. Once the corners are obtained for each page, the geometrical distortion is rectified by a perspective transformation method. When two adjacent pages are coupled, as often occurs in scanned and camera-captured comics, they are separated by a simple low-energy point method. The distortion-reducing performance is measured in terms of identification accuracy using the Hamming distance. In a simulation study, our proposed scheme improved the average fingerprint identification accuracy by more than 30 percentage point for single-page comics and by 28 percentage point for double-page comics, relative to conventional methods. We also analyzed the execution time of distortion reduction and page separation in low-, medium- and high-resolution images. Even for large images, the average processing time of our scheme was within 2 s, which is a sufficiently short time for commercial applications.

Discriminative feature learning‐based pixel difference representation for facial expression recognition

Recently, researchers have proposed different feature descriptors to achieve robust performance for facial expression recognition (FER). However, finding a discriminative feature descriptor remains one of the critical tasks. In this paper, we propose a discriminative feature learning scheme to improve the representation power of expressions. First, we obtain a discriminative feature matrix (DFM) based pixel difference representation. Subsequently, all DFMs corresponding to the training samples are used to construct a discriminative feature dictionary (DFD). Next, DFD is projected on a vertical two‐dimensional linear discriminant analysis in direction (V‐2DLDA) space to compute between and within‐class scatter because V‐2DLDA works well with the DFD in matrix representation and achieves good efficiency. Finally, nearest neighbor (NN) classifier is used to determine the labels of the query samples. DFD represents the local feature changes that are robust to the expression, illumination et al. Besides, we exploit V‐2DLDA to find an optimal projection matrix since it not only protects the discriminative features but reduces the dimensions. The proposed method achieves satisfying recognition results, reaching accuracy rates as high as 91.87% on CK+ database, 82.24% on KDEF database, and 78.94% on CMU Multi‐PIE database in the LOSO scenario, which perform better than other comparison methods.

A Novel Segment-Based Data Transfer Scheme for Low-Power CMOS Surveillance Image Sensors

Smart image sensors with low data rate output are well fitted for security and surveillance tasks, since at lower data rates, power consumption is reduced and the image sensor can be operated with limited energy resources such as solar panels. In this paper, a new data transfer scheme is presented to reduce the data rate of the pixels which have undergone value change. Although different pixel difference detecting architectures have been previously reported but it is shown that the given method is more effective in terms of power dissipation and data transfer rate reduction. The proposed architecture is evaluated as a [Formula: see text]-pixel sensor in a standard CMOS technology and comparison with other data transfer approaches is performed in the same process and configuration.

Exploring metal artifact reduction using dual-energy CT with pre-metal and post-metal implant cadaver comparison: are implant specific protocols needed?

ObjectiveTo quantify and optimize metal artifact reduction using virtual monochromatic dual-energy CT for different metal implants compared to non-metal reference scans.MethodsDual-energy CT scans of a pair of human cadaver limbs were acquired before and after implanting a titanium tibia plate, a stainless-steel tibia plate and a titanium intramedullary nail respectively. Virtual monochromatic images were analyzed from 70 to 190 keV. Region-of-interest (ROI), used to determine fluctuations and inaccuracies in CT numbers of soft tissues and bone, were placed in muscle, fat, cortical bone and intramedullary tibia canal.ResultsThe stainless-steel implant resulted in more pronounced metal artifacts compared to both titanium implants. CT number inaccuracies in 70 keV reference images were minimized at 130, 180 and 190 keV for the titanium tibia plate, stainless-steel tibia plate and titanium intramedullary nail respectively. Noise, measured as the standard deviation of pixels within a ROI, was minimized at 130, 150 and 140 keV for the titanium tibia plate, stainless-steel tibia plate and titanium intramedullary nail respectively.ConclusionTailoring dual-energy CT protocols using implant specific virtual monochromatic images minimizes fluctuations and inaccuracies in CT numbers in bone and soft tissues compared to non-metal reference scans.

A hybrid approach for face alignment

Face alignment has been an indispensable procedure in face application. It is still a challenge to locate facial landmarks in unconstrained scene. In this paper, we propose an algorithm to perform accurately face alignment. A method based on human retinal information processing principle is proposed to enhance the images and remove the illumination noise. According to the image’s locality principle, the distance constrains is imposed on the pixel difference features around the landmark to achieve robustness. And then, the random forest is used to map the pixel difference features to local binary features. The obtained local binary features are used to jointly learn a linear regression for the final output. In addition, the inherent data structure is utilized to reduce the computational burden when preform maximum variance reduction on the split node in random forest. Extensive experiments on public datasets show that the proposed approach can locate facial landmarks accurately and rapidly.

화소차이에 따른 적응적 임계값 결정방법을 융합한 동작검출 알고리즘

화소차이에 따른 적응적 임계값 결정방법을 융합한 동작검출 알고리즘

Context model based edge preservation filter for impulse noise removal

In this article, a new edge preserving contextual model based image restoration technique is proposed for images affected by impulse noise. The proposed restoration technique consists of two stages: noisy pixel identification and restoration. Center sliding window is considered as current processing pixel for both noisy pixel identification and restoration. In the first stage of the proposed technique, we follow an absolute directional difference of the neighborhood pixels to identify the pixels those are affected by impulse noise. We propose an edge preserving contextual model to restore the noisy pixels. The noise correction stage of the proposed scheme depends on the context model of the noise-free pixels in the selected window. The parameters of the contextual model are obtained using a Gaussian kernel. The proposed algorithm is tested on nine benchmark test images. The evaluation of the proposed algorithm is carried out by comparing it against nine competitive state-of-the-art algorithms for impulse noise removal. The proposed algorithm is evaluated using Peak Signal to Noise Ratio (PSNR), Mean Structural Similarity Index (MSSIM), Non-shifted Edge Ratio (NSER) and Correlation Factor (CF) performance measures. Experimental results corroborate that the proposed algorithm provides better performance than the existing state-of-art impulse denoising methods.

Fusing Local and Global Features for High-Resolution Scene Classification

In this paper, a fused global saliency-based multiscale multiresolution multistructure local binary pattern (salM 3LBP) feature and local codebookless model (CLM) feature is proposed for high-resolution image scene classification. First, two different but complementary types of descriptors (pixel intensities and differences) are developed to extract global features, characterizing the dominant spatial features in multiple scale, multiple resolution, and multiple structure manner. The micro/macrostructure information and rotation invariance are guaranteed in the global feature extraction process. For dense local feature extraction, CLM is utilized to model local enrichment scale invariant feature transform descriptor and dimension reduction is conducted via joint low-rank learning with support vector machine. Finally, a fused feature representation between salM3LBP and CLM as the scene descriptor to train a kernel-based extreme learning machine for scene classification is presented. The proposed approach is extensively evaluated on three challenging benchmark scene datasets (the 21-class land-use scene, 19-class satellite scene, and a newly available 30-class aerial scene), and the experimental results show that the proposed approach leads to superior classification performance compared with the state-of-the-art classification methods.

Full-Field Calibration of Color Camera Chromatic Aberration using Absolute Phase Maps.

The refractive index of a lens varies for different wavelengths of light, and thus the same incident light with different wavelengths has different outgoing light. This characteristic of lenses causes images captured by a color camera to display chromatic aberration (CA), which seriously reduces image quality. Based on an analysis of the distribution of CA, a full-field calibration method based on absolute phase maps is proposed in this paper. Red, green, and blue closed sinusoidal fringe patterns are generated, consecutively displayed on an LCD (liquid crystal display), and captured by a color camera from the front viewpoint. The phase information of each color fringe is obtained using a four-step phase-shifting algorithm and optimum fringe number selection method. CA causes the unwrapped phase of the three channels to differ. These pixel deviations can be computed by comparing the unwrapped phase data of the red, blue, and green channels in polar coordinates. CA calibration is accomplished in Cartesian coordinates. The systematic errors introduced by the LCD are analyzed and corrected. Simulated results show the validity of the proposed method and experimental results demonstrate that the proposed full-field calibration method based on absolute phase maps will be useful for practical software-based CA calibration.

Evidence for compact binary systems around Kepler red giants

We present an analysis of 168 oscillating red giants from NASA's $Kepler$ mission that exhibit anomalous peaks in their Fourier amplitude spectra. These peaks result from ellipsoidal variations which are indicative of binary star systems, at frequencies such that the orbit of any stellar companion would be within the convective envelope of the red giant. Alternatively, the observed phenomenon may be due to a close binary orbiting a red giant in a triple system, or chance alignments of foreground or background binary systems contaminating the target pixel aperture. We identify 87 stars in the sample as chance alignments using a combination of pixel Fourier analysis and difference imaging. We find that in the remaining 81 cases the anomalous peaks are indistinguishable from the target star to within 4$''$, suggesting a physical association. We examine a Galaxia model of the $Kepler$ field of view to estimate background star counts and find that it is highly unlikely that all targets can be explained by chance alignments. From this, we conclude that these stars may comprise a population of physically associated systems.

A method to extract image noise level from patient images in CT.

Image quality in computed tomography (CT) is usually not quantified in real time from patient scans. Rather, phantom scans are employed to measure the contrast, noise, and spatial resolution properties of a given system. This approach, however, is not ideal because many factors are difficult to represent using phantoms like variations in patient size, composition, and position within the gantry. Several methods for measuring the noise found within patient scans have been proposed. All of these methods rely on segmenting out relatively uniform regions within the patient in order to differentiate between morphological variations and pixel differences due to photon noise. In order to avoid this segmentation step, we propose to analyze the noise signal in the air surrounding the patient. We will demonstrate that the air signal surrounding the patient acts as a surrogate for the noise within the contours of the patient. Our work builds off the global noise index (GNI) method. In the GNI method, adjacent axial CT image slices are subtracted to remove the majority of the morphological variations in the data. Remaining morphological variations are removed with an edge-finding algorithm. Then the image is divided up into small regions of interest (ROI) and the pixel standard deviation is computed for each ROI. Only those ROIs not containing bone or air are then used to make a histogram of the standard deviation within the image. The mode of this histogram is referred to as the GNI. We refer to this as the traditional GNI (tradGNI). Our modification to this workflow is to apply this metric to just the air signal surrounding the patient. We evaluate the correlation between using the air signal and the in-phantom metric value for titration over: dose level, image slice thickness, kernel sharpness, iterative reconstruction level, model based iterative reconstruction, reconstruction field of view, and reconstruction interval. 373 patient abdomen pelvis exams were collected and the air based noise estimation metrics applied. Liver standard deviation values were measured on 40 of the patients and correlated with air GNI calculations. Our results show excellent linear correlation (R2 > 0.99) between the tradGNI being applied inside and outside of a phantom object. Our results are also shown to still be predictive of the noise under all scan parameters studied, including iterative and model based reconstruction. Fits of tradGNI versus dose level exhibited the expected square root dependence. Air based GNI metrics were predictive of human patient noise level and also correlated with manual liver standard deviation measurement (R2 = 0.73). Our results demonstrate the signal in the air surrounding an imaging object can accurately be used as a surrogate for the image noise within the object. Our method should enable faster and more robust patient specific image quality assessment due to the lack of the need to segment noise from morphological variations within a patient.

Magnetic resonance based texture parameters as potential imaging biomarkers for predicting long‐term survival in locally advanced rectal cancer treated by chemoradiotherapy

The study aimed to investigate whether textural features of rectal cancer on MRI can predict long-term survival in patients treated with long-course chemoradiotherapy. Textural analysis (TA) using a filtration-histogram technique of T2-weighted pre- and 6-week post-chemoradiotherapy MRI was undertaken using TexRAD, a proprietary software algorithm. Regions of interest enclosing the largest cross-sectional area of the tumour were manually delineated on the axial images and the filtration step extracted features at different anatomical scales (fine, medium and coarse) followed by quantification of statistical features [mean intensity, standard deviation, entropy, skewness, kurtosis and mean of positive pixels (MPP)] using histogram analysis. Cox multiple regression analysis determined which univariate features including textural, radiological and histological independently predicted overall survival (OS), disease-free survival (DFS) and recurrence-free survival (RFS). MPP [fine texture, hazard ratio (HR) 6.9, 95% CI: 2.43-19.55, P<0.001], mean (medium texture, HR 5.6, 95% CI: 1.4-21.7, P=0.007) and extramural venous invasion (EMVI) on MRI (HR 2.96, 95% CI: 1.04-8.37, P=0.041) independently predicted OS while mean (medium texture, HR 4.53, 95% CI: 1.58-12.94, P=0.003), MPP (fine texture, HR 3.36, 95% CI: 1.36-8.31, P=0.008) and threatened circumferential resection margin (CRM) on MRI (HR 3.1, 95% CI: 1.01-9.46, P=0.046) predicted DFS. For OS, EMVI on MRI (HR 4.23, 95% CI: 1.41-12.69, P=0.01) and for DFS kurtosis (medium texture, HR 3.97, 95% CI: 1.44-10.94, P=0.007) and CRM involvement on MRI (HR 3.36, 95% CI: 1.21-9.32, P=0.02) were the independent post-treatment factors. Only TA independently predicted RFS on pre- or post-treatment analyses. MR based TA of rectal cancers can predict outcome before undergoing surgery and could potentially select patients for individualized therapy.

Extended three-dimensional rotation invariant local binary patterns

This paper presents a new set of three-dimensional rotation invariant texture descriptors based on the well-known local binary patterns (LBPs). In the approach proposed here, we extend an existing three-dimensional LBP based on the region growing algorithm using existing features developed exquisitely for two-dimensional LBPs (pixel intensities and differences). We have conducted experiments on a synthetic dataset of three-dimensional randomly rotated texture images in order to evaluate the discriminatory power and the rotation invariant properties of our descriptors as well as those of other two-dimensional and three-dimensional texture descriptors. Our results demonstrate the effectiveness of the extended LBPs and improvements against other state-of-the-art hand-crafted three-dimensional texture descriptors on this dataset. Furthermore, we prove that the extended LBPs can be used in medical datasets to discriminate between MR images of oxygenated and non-oxygenated brain tissues of newborn babies.

New dissimilarity measures for image phylogeny reconstruction

Image phylogeny is the problem of reconstructing the structure that represents the history of generation of semantically similar images (e.g., near-duplicate images). Typical image phylogeny approaches break the problem into two steps: (1) estimating the dissimilarity between each pair of images and (2) reconstructing the phylogeny structure. Given that the dissimilarity calculation directly impacts the phylogeny reconstruction, in this paper, we propose new approaches to the standard formulation of the dissimilarity measure employed in image phylogeny, aiming at improving the reconstruction of the tree structure that represents the generational relationships between semantically similar images. These new formulations exploit a different method of color adjustment, local gradients to estimate pixel differences and mutual information as a similarity measure. The results obtained with the proposed formulation remarkably outperform the existing counterparts in the literature, allowing a much better analysis of the kinship relationships in a set of images, allowing for more accurate deployment of phylogeny solutions to tackle traitor tracing, copyright enforcement and digital forensics problems.

Sound-Event Classification Using Robust Texture Features for Robot Hearing

Sound-event classification often utilizes time–frequency analysis, which produces an image-like spectrogram. Recent approaches such as spectrogram image features and subband power distribution image features extract the image local statistics such as mean and variance from the spectrogram. They have demonstrated good performance. However, we argue that such simple image statistics cannot well capture the complex texture details of the spectrogram. Thus, we propose to extract the local binary pattern (LBP) from the logarithm of the Gammatone-like spectrogram. However, the LBP feature is sensitive to noise. After analyzing the spectrograms of sound events and the audio noise, we find that the magnitude of pixel differences, which is discarded by the LBP feature, carries important information for sound-event classification. We thus propose a multichannel LBP feature via pixel difference quantization to improve the robustness to the audio noise. In view of the differences between spectrograms and natural images, and the reliability issues of LBP features, we propose two projection-based LBP features to better capture the texture information of the spectrogram. To validate the proposed multichannel projection-based LBP features for robot hearing, we have built a new sound-event classification database, the NTU-SEC database, in the context of social interaction between human and robot. It is publicly available to promote research on sound-event classification in a social context. The proposed approaches are compared with the state of the art on the RWCP database and the NTU-SEC database. They consistently demonstrate superior performance under various noise conditions.

눈 깜박임 화소 값 기반의 안면과 홍채영역 영상인식용 모듈설계

본 논문에서는 홍채정보로 개인인증을 위한 USB-OTG(Uiversal Serial Bus On-the-go) 영상인식 모듈을 설계한다. 개인인증을 위해 사용자가 스마트 폰 버튼을 누를 필요가 없도록 스마트 기기를 안면주위의 여러 장의 안면영상을 획득 후, 눈 깜박임에 의한 화소 값 차로 안면과 홍채영역을 검색하는 영상인식 알고리듬을 제안한다. 본 연구에서는 인접한 눈을 뜬 영상과 눈을 감은 영상을 감지한 안면과 홍채 영상의 프레임 화소 값의 차이를 사용한다. 또한, 홍채 영역분할에 의한 동공과 홍채영역 위치를 빠르게 찾을 수 방법을 활용한다. 제안한 빠른 홍채영역의 위치탐색은 눈 영역의 적정한 그리드 크기에 의해 결정할 수 있다. 안면과 홍채영역의 제한된 영역을 탐색하는 홍채인식 카메라 모듈의 USB-OTG 인터페이스 통한 인접영상의 프레임 차이에 의해 검출할 수 있도록 설계하였다. 이로서 스마트 디바이스 사용자가 홍채 인식을 위해 눈을 깜빡이지 않고 대기해야 하는 불편함을 제거함으로써 사용자 편의성을 증대시킬 것으로 기대한다. In this paper, an USB-OTG (Uiversal Serial Bus On-the-go) interface module was designed with the iris information for personal identification. The image recognition algorithm which was searching face and iris areas, was proposed with pixel differences from eye blinking after several facial images were captured and then detected without any activities like as pressing the button of smart phone. The region of pupil and iris could be fast involved with the proper iris area segmentation from the pixel value calculation of frame difference among the images which were detected with two adjacent open-eye and close-eye pictures. This proposed iris recognition could be fast processed with the proper grid size of the eye region, and designed with the frame difference between the adjacent images from the USB-OTG interface with this camera module with the restrict of searching area in face and iris location. As a result, the detection time of iris location can be reduced, and this module can be expected with eliminating the standby time of eye-open.

Local derivative radial patterns: A new texture descriptor for content-based image retrieval

In this paper, we propose a novel local pattern descriptor called Local Derivative Radial Pattern (LDRP) for texture representation in content-based image retrieval. All prior local patterns are based on gray-level difference of pixels located in a square or circle. Since many of the actual textures can be represented by intensity relationship of pixels along a line, these methods do not have a suitable ability to represent texture information. In prior methods, difference between referenced pixel and its adjacent pixel is encoded with two, three or four values which leads to information loss of the image. The proposed LDRP is based on gray-level difference of pixels along a line and their weighted combinations. In addition, multi-level coding in different directions is used instead of binary coding. The performance of the proposed method is compared with prior methods including local binary pattern (LBP), local ternary pattern (LTP), local derivative pattern (LDP), local tetra pattern (LTrP) and local vector pattern (LVP). The proposed LDRP outperforms all mentioned prior methods by at least 3.82% and 5.17% in terms of average precision on Brodatz and VisTex databases, respectively.

Computer-Based Algorithmic Determination of Muscle Movement Onset Using M-Mode Ultrasonography

The study purpose was to evaluate the use of computer-automated algorithms as a replacement for subjective, visual determination of muscle contraction onset using M-mode ultrasonography. Biceps and quadriceps contraction images were analyzed visually and with three different classes of algorithms: pixel standard deviation (SD), high-pass filter and Teager Kaiser energy operator transformation. Algorithmic parameters and muscle onset threshold criteria were systematically varied within each class of algorithm. Linear relationships and agreements between computed and visual muscle onset were calculated. The top algorithms were high-pass filtered with a 30 Hz cutoff frequency and 20 SD above baseline, Teager Kaiser energy operator transformation with a 1200 absolute SD above baseline and SD at 10% pixel deviation with intra-class correlation coefficients (mean difference) of 0.74 (37.7 ms), 0.80 (61.8 ms) and 0.72 (109.8 ms), respectively. The results suggest that computer automated determination using high-pass filtering is a potential objective alternative to visual determination in human movement science.