运用肤色信息和模板匹配的彩色人脸检测

人脸是一个复杂的模式,在图像中自动地对其进行定位和分割是进行人脸识别的第一步。本文提出一种运用肤色信息和模板匹配的人脸检测方法。该方法先进行肤色分割,然后对每一个人脸候选区域进行形状比例的分析,最后进行模板匹配。实验结果表明,该方法对任意背景下,任意姿态及任意数目的人脸检测非常有效。     

分水岭变换和统计区域合并的图像分割算法研究

提出了一种基于分水岭变换和统计区域合并的图像分割方法.该方法综合利用高斯低通滤波、分水岭变换和统计区域合并,先对原始图像提取分割标记,然后利用Meyer分水岭变换对标记分水岭进行分割,最后利用概率统计的方法对过分割区域进行合并.该算法通过调节尺度参数可以实现由粗到细(coarse-to-fine)的分割.实验结果表明,这种简单可行的算法在分割噪声图像时依然有良好的效果,具有较强的鲁棒性.     

Automatic merging of tetrahedral meshes

A generic algorithm is proposed to merge arbitrary solid tetrahedral meshes automatically into one single valid finite element mesh. The intersection segments in the form of distinct nonoverlapping loops between the boundary surfaces of the given solid objects are determined by the robust neighbor tracing technique. Each intersected triangle on the boundary surface will be triangulated to incorporate the intersection segments onto the boundary surface of the objects. The tetrahedra on the boundary surface associated with the intersected triangular facets are each divided into as many tetrahedra as the number of subtriangles on the triangulated facet. There is a natural partition of the boundary surfaces of the solid objects by the intersection loops into a number of zones. Volumes of intersection can now be identified by collected bounding surfaces from the surface patches of the partition. Whereas mesh compatibility has already been established on the boundary of the solid objects, mesh compatibility has yet to be restored on the bounding surfaces of the regions of intersection. Tetrahedra intersected by the cut surfaces are removed, and new tetrahedra can be generated to fill the volumes bounded by the cut surfaces and the portion of cavity boundary connected to the cut surfaces to restore mesh compatibility at the cut surfaces. Upon restoring compatibility on the bounding surfaces of the regions of intersection, the objects are ready to be merged together as all regions of intersection can be detached freely from the objects. All operations, besides the determination of intersections structurally in the form of loops, are virtually topological, and no parameter and tolerance is needed in the entire merging process. Examples are presented to show the steps and the details of the mesh merging procedure. Copyright © 2012 John Wiley & Sons, Ltd.     

A novel energy‐based approach for merging finite elements

A novel approach for merging two intersecting finite elements is presented and demonstrated. The solution mimics concepts from biology and uses principles rooted in continuum mechanics. The problem of attaching (or merging) two coincident finite elements is common when using the plastering technique as part of the advancing front method. This problem is particularly challenging for 3‐D meshes of non‐convex shapes. Some automatic meshing methods require portions of the partially formed mesh to coincide and merge. This problem is generally solved with heuristic rules, which lack generality, and may have difficulties with unforeseen situations. The problem of merging two overlapping polyhedra may also appear in other applications such as computer graphics and CAD software. A new approach to address the problem of merging is presented here. This solution does not utilize heuristic rules, but rather uses an approach based on minimization of strain energy. A fully automatic merging routine has been created that can address, in an optimum way, any situation of two nearby or overlapping elements that are to be merged. This approach, with minor adjustments, is suitable for most types of 3‐D elements. Copyright © 2010 John Wiley & Sons, Ltd.     

A dynamic threshold-based local mesh ternary pattern technique for biomedical image retrieval

Many content-based image retrieval techniques like local binary pattern (LBP), local ternary pattern (LTP), local mesh peak valley edge pattern (LMePVEP), local mesh ternary pattern (LMeTerP), etc. extract texture features of an image for retrieval purposes. These techniques use fixed threshold to encode the input image and selection of such threshold value is not guided, that is, a chosen threshold may not be optimal for all images in the database. Moreover the performance of these texture-based static threshold algorithms also decreases if the input images are noisy. In this paper, a dynamic threshold value-based local mesh ternary pattern method is proposed in which the threshold value is chosen from the neighborhood of a central pixel using median of all pixels. The proposed modification reduces the overall effect of noise component and thereby improves the average retrieval rate (ARR) and average retrieval precision (ARP) of the original technique. The proposed modified technique has been compared with five other image retrieval approaches to prove its worthiness - the original local mesh ternary pattern technique (LMeTerP), a local ternary pattern technique (LTP), a Best ensemble technique, a multi-label classification CNN model and a CNN-based model of the proposed approach using a VIA ELCAP lung database and an Emphysema database. An improvement of 3.92% in ARR and 2.53% in ARP is observed over the basic local mesh ternary pattern method. Further the proposed modification has been combined with CNN concept and its results have also been analyzed.     

A novel hybrid artificial neural network technique for the early skin cancer diagnosis using color space conversions of original images

In this study, an innovative hybrid machine learning-technique is used for the early skin cancer diagnosis fusing Convolutional Neural Network and Multilayer Perceptron to analyze images and information related to the skin cancer. This information is extracted manually after applying different color space conversions on the original images for better screening of the lesions. The proposed architecture is compared with standalone architecture in addition to some other techniques by commonly used evaluation metrics. HAM10000 dataset is used for training and testing as this data contain seven different skin lesions. The novelty of the proposed hybrid model is the structure of the network which handles structured data (patients' metadata and other useful features from different color spaces related to the illumination, energy, darkness, etc.) and unstructured data (images). The results show an overall 86%, 95% top-1 and top-2 accuracy respectively, and 96% area under the curve for the seven classes. The study demonstrates the superiority of the proposed hybrid model with a 2% improvement in the accuracy over the standalone model and a promising behavior as compared to the ensemble techniques. The follow-up research will include more patient data to develop a skin cancer detection device.     

Application of the image processing technique in identifying the particle dispersion from a centrifugal fertilizer spreader

Particle dispersion in the vicinity of an agricultural fertilizer spreader is difficult to capture due to the rapid particle traveling motion. This paper introduced a granule impact indentation-based technique to simultaneously record the two-dimensional particle dispersion from a spinning concave disc type of spreader. A Nitrogen-Phosphorus-Potassium (NPK) type of fertilizer was utilized to induce indentations on aluminum foils placed on the wall panels confining an experimental spreader system. Subsequently, an image processing technique which is comprised of the multicolor edge detection, the curve closing, and the region merging techniques was purposely developed to automatically identify and locate the granule impacts on the sampled foil digital images. Overlapping impacts were characterized based on the granulometry of the fertilizer sample. The reconstructed particle dispersion pattern using the image processing method showed good agreement with the experimental observations. The outcome of this research enabled a fast and effective method for quantitatively assessing the particle distribution for a specific fertilizer spreader.     

向量空间的彩色图像边缘检测

边缘检测是图像处理的重要组成部分之一。拓展小波变换为向量空间形式,并提出了基于向量小波变换的彩色图像边缘检测方法。根据小波变换的特性,检测模值局部极大点作为边缘点。引入大小自适应的卷积模板,小波变换尺度为 2j时,大小为(4×2j-1)×(4×2j-1),使计算量得到有效控制。实验结果证明,该方法继承了小波变换的性能,并充分利用了彩色信息,能完全地检测出彩色图像边缘。     

Characterization of trichromatic color cameras by using a new multispectral imaging technique

We investigate methods for the recovery of reflectance spectra from the responses of trichromatic camera systems and the application of these methods to the problem of camera characterization. The recovery of reflectance from colorimetric data is an ill-posed problem, and a unique solution requires additional constraints. We introduce a novel method for reflectance recovery that finds the smoothest spectrum consistent with both the colorimetric data and a linear model of reflectance. Four multispectral methods were tested using data from a real trichromatic camera system. The new method gave the lowest maximum colorimetric error in terms of camera characterization with test data that were independent of the training data. However, the average colorimetric performances of the four multispectral methods were statistically indistinguishable from each other but were significantly worse than conventional methods for camera characterization such as polynomial transforms.     

A novel automated depression detection technique using text transcript

Depression is one of the most common mental illnesses, impacting billions of people worldwide. The lack of existing resources is impeding the country's economic prosperity. As a result, new approaches for detecting and treating mental diseases as well as reaching out to individuals are required so that people can overcome their daily challenges and become more productive. An automated depression detection system can greatly aid in clinical findings and early treatment of depression. Automatic detection, like in a clinical interview can be derived from various modalities that include video, audio, and text. Among these modalities, audio characteristics are the most commonly researched while text elements are seldom investigated. In the light of the above, this paper proposes a novel automated depression identification approach based on linguistic material gathered from patient interviews. The focus is to enhance both the accuracy and efficiency of detection. The proposed approach is made up of two parts: a Bidirectional Gated Recurrent Unit (BGRU) network for dealing with linguistic information and a fully coupled network that integrates the model outputs to measure the depressed state. The proposed approach is validated using Distress Analysis Interview Corpus-Wizard-of-Oz interviews dataset. To evaluate the performance precision, recall, and F1 score are computed using the proposed approach and then the comparative analysis is done with the existing approaches. The proposed approach yielded an F1 score of 0.92, indicating the existence of depression as well as the projected severity level. It is realized from the generated results that the proposed approach has outperformed the previous ones. The proposed approach can not only automatically assess the severity of depression but also enhances both the accuracy and efficiency of detection. The proposed approach indicates the feasibility of BGRU over Long Short Term Memory by achieving exceptional results for recognition of depression.     

Improvement of color reproduction in color digital holography by using spectral estimation technique

We propose a color digital holography by using spectral estimation technique to improve the color reproduction of objects. In conventional color digital holography, there is insufficient spectral information in holograms, and the color of the reconstructed images depend on only reflectances at three discrete wavelengths used in the recording of holograms. Therefore the color-composite image of the three reconstructed images is not accurate in color reproduction. However, in our proposed method, the spectral estimation technique was applied, which has been reported in multispectral imaging. According to the spectral estimation technique, the continuous spectrum of object can be estimated and the color reproduction is improved. The effectiveness of the proposed method was confirmed by a numerical simulation and an experiment, and, in the results, the average color differences are decreased from 35.81 to 7.88 and from 43.60 to 25.28, respectively.     

甲醛分析仪现场检测技术

采用扩散管动态配气装置发生甲醛标准气体，对BH-1型甲醛快速分析仪的各项性能指标（响应时间，稳定性）进行了测试，确定了分析条件（温度，相对湿度，风速）对仪器的影响，结果显示快速分析仪测定甲醛气体浓度的相对误差小于10％。研究表明经定期校准的甲醛快速分析仪用于室内甲醛气体污染的检测可获得可靠的结果。     

基于肤色特征和动态聚类的彩色人脸检测

在人类视觉机制和肤色聚类特性的基础上,提出了一种复杂背景下人脸检测方法。该方法采用K-均值动态聚类分析算法,利用人类肤色特征在输入图像中检测包含人脸的似人脸区作为候选人脸,再用同样的方法对候选人脸区域进行扫描,得到真正的人脸。实验结果表明,该方法的正确检出率达到84%,受背景、光照、角度、姿态的影响很小,具有较好的鲁棒性。     

复合材料B基准值统计的多环境样本合并方法

为缩减获得复合材料设计许用值的试验成本,首先,介绍了复合材料手册-17-G最新推荐的一种基于回归分析的合并多环境样本的B基准值统计方法,重点阐述了在批间变异检验、环境样本间的方差等同性检验及正态性检验中加入工程经验判断的合理性及适用条件;然后,根据该多环境样本合并方法,设计开发了B基准值统计分析软件;最后,采用该软件对江苏恒神股份有限公司部分复合材料的力学性能数据进行了统计分析。结果表明:与传统的单环境样本统计程序相比,该多环境样本合并方法不仅显著提高了B基准值,而且克服了传统统计程序中对样本数量及批次数的限制。所得结论表明该方法具有较高的可操作性,适用于在缩减样本的情况下统计分析复合材料B基准值。      

A novel region-based multimodal image fusion technique using improved dictionary learning

Recently, the sparse representation (SR) based algorithms have gained much attention from the researchers in the area of image fusion (IF). The building of a compact discriminative dictionary plays a vital role in the sparse-based IF techniques. In this context, an efficient multimodal IF method based on improved dictionary learning is investigated. The key contributions of this paper are: (a) An improved KSVD algorithm is suggested for the dictionary learning process, (b) to reduce the computational time, only the informative patches are selected using energy feature, and (c) a novel region-based fusion scheme is suggested for the first time for the problem on hand. The suggested technique is tested with a number of multimodal images from Harvard Medical School brain database. The results are compared with state-of-the-art multiscale transform-based methods and modified SR-based methods. Unlike earlier methods, our proposed technique generates an adaptive dictionary through selection of informative patches only. This results in a compact dictionary with improved computational efficiency. The experimental results reveal that our approach outperforms other methods. The potential application of the suggested method could be in pathological images for follow-up study and better treatment planning.     

A computer-controlled ultrasound pulser-receiver system for transskull fluid detection using a shear wave transmission technique

The purpose of this study was to evaluate the performance of a computer-controlled ultrasound pulser-receiver system incorporating a shear mode technique for transskull fluid detection. The presence of fluid in the sinuses of an ex vivo human skull was examined using a pulse-echo method by transmitting an ultrasound beam through the maxilla bone toward the back wall on the other side of the sinus cavity. The pulser was programmed to generate bipolar pulse trains with 5 cycles at a frequency of 1 MHz, repetition frequency of about 20 Hz, and amplitude of 100 V to drive a 1-MHz piezoelectric transducer. Shear and longitudinal waves in the maxilla bone were produced by adjusting the bone surface incident angle to 45 degrees and 0 degrees, respectively. Computer tomography (CT) scans of the skull were performed to verify the ultrasound experiment. Using the shear mode technique, the echo waveform clearly distinguishes the presence of fluid, and the estimated distance of the ultrasound traveled in the sinus is consistent with the measurement from the CT images. Contrarily, using the longitudinal mode, no detectable back wall echo was observed under the same conditions. As a conclusion, this study demonstrated that the proposed pulser-receiver system with the shear mode technique is promising for transskull fluid detecting, such as mucus in a sinus.     

Emboli detection using the Doppler ultrasound technique

Embolic detection is very important to the early diagnosis of vessel disease. The Doppler ultrasound technique is one of the common methods to detect the emboli non-invasively. When the emboli pass through the sample volume of the Doppler ultrasound instrument, there exist high intensity transient Doppler signals. Thus the emboli can be detected directly from the variation of Doppler signal amplitude. Since there may be some disturbance in the system, this general detection method has great limitation. To improve the accuracy of emboli auto-detection, several novel methods are studied to obtain the sensitive characteristic of the emboli signals using the new signal processing theories.     

利用肤色分割和对称变换实现鲁棒人脸特征定位

提出一种将肤色分割与灰度图像对称变换相结合的有效的人脸特征定位方法.为降低背景和人脸姿态的影响,在传统人脸肤色分割基础上,利用Blob分析及椭圆拟和方法判断并剔除无效区域,并将待选人脸区域初步校正为竖直;直接利用色度信息在待选人脸区域中分割并检测嘴唇;接着,采用灰度图像的对称变换检测双眼的待选位置;以嘴唇和待选双眼的位置为依据,提出了基于人脸几何模型知识的双眼匹配代价函数并利用组合优化方法检测真实双眼;最后,对人脸上其它器官的进行精确定位.实验结果表明,该方法对于头发遮挡、大倾斜角以及眼镜干扰具有较强的鲁棒性.     

二值模式的一种快速细化方法

本文提出了一种二值模式的快速细化方法,该算法能有效地抑制模式上的“毛刺”,快速提取保持原始图象连通性和拓扑性的骨架,它的处理速度和效果明显优于文献[1]中所述的算法。文中我们给出了两种算法对汉字“国”和“E”,“H”,“G”等字母处理后的结果。     

Diagnosis of a dermatological lesion using intelligent feature selection technique

The content-based image retrieval (CBIR) in dermatological diagnosis context, the information matching is the major concern in terms of feature vector-based classification. The discrimination of the feature vector leads to better classification as well as retrieval rate. Better retrieval results help the dermatologist to improve the diagnosis. In this paper, we proposed a support vector machine weight map (SVM W-Map)-based feature selection along with multi-class particle swarm optimization (PSO) presented for multi-class dermatological imaging dataset. The performance of the system was tested on a dataset including 1450 images and obtained 99.7% for specificity and 95.89% for sensitivity. The analysis and evaluations of results show that the proposed system has higher diagnosis ability when compared with other works.