Multifocus image fusion scheme based on feature contrast in lifting stationary wavelet domain

A novel multifocus image fusion method based on lifting stationary wavelet transform (LSWT) is proposed. The selection principles, namely fusion rules of different subband coefficients, are discussed in detail. Local feature contrast is presented according to the human vision system (HVS), which is highly sensitive to the local image contrast level. Then, the fusion rule for the low-frequency subband coefficients fusion is introduced. To choose the high frequency subband coefficients, another local feature contrast is developed according to the human vision which is often sensitive to edges and directional features, but insensitive to real luminance at independent positions. Then, a novel fusion rule is proposed for fusion of the high frequency subband coefficients. Experimental results demonstrate that the proposed image fusion method is effective and can provide better performance in fusing multifocus images than the traditional contrast-based image fusion algorithms in term of informal visual inspection and objective criteria in multi-focus image fusion.     

Image denoising based on adaptive graph regularization

Adaptive regularization can select different parameters based on the features of local areas in an image, which can differentiate the edges and noise in an image flexibly. An adaptive graph regularization is proposed based on graph spectral theory and adaptive regularization, which uses the Non local means to generate the weighting function of graph. The adaptive graph regularization equation is used to filter the noisy image. Simulation results show that the proposed method can effectively remove the noise and is superior to other graph theory based partial differential equation methods.     

Multi scale and multi orientation features for image de noising

An adaptive algorithm for image de noising is proposed based on the multi scale and multi orientation features. The coefficients in different scales and different directions are obtained by image decomposition using the nonsubsampled contourlet transform. Then thresholds functions are adaptively set with these coefficients. The texture of the image information is introduced by using the mean of decomposition scale and the energy of regional. The greater the energy, the more information of the texture while the same decomposition scales, the smaller the threshold is set. On the contrary, the greater the threshold is set. After the de noising and then reconstruction of these coefficients, image de noising is implemented. Compare to the wavelet transform threshold and contourlet transform threshold, the nonsubsampled contourlet transform pick up the image detail better and improve the quality of the image.     

Image edge detection algorithm in fuzzy domain

An image edge detection algorithm in fuzzy domain is proposed, which combines adaptive fuzzy enhancement and multi direction fuzzy morphology to detect the edges of fuzzy image. The adaptive fuzzy enhancement method enhances the fuzzy image within blocks with sliding windows to avoid losses of the real edges resulting from enhancing with single threshold for the whole image and lead strong adaptive ability to image region variance. The multi direction fuzzy mathematical morphology operates on the enhanced fuzzy image with structure elements of multiple different directions to extract the real edges with directionality and restrain non directional noise. Experiments show the algorithm can detect fuzzy image edge effectively with strong antinoise ability.     

An improved morphological edge detection algorithm of medical image

Medical images usually contain much noise which affects the edge detection accuracy. Focusing on this problem, based on the edge detection operator in mathematical morphology, an improved edge detection algorithm is presented by combining the features of the multi structure elements and the multi scale edge detection algorithm. The algorithm performs opening and closing operations on the data with the alternative sequence filters and the structure elements. The weighting operation is applied with different weight coefficients for horizontal, vertical and diagonal directions, while the edge detection operator with dilation type is calculated to obtain the improved edge detection algorithm. The steps of the algorithm are described. The algorithm is used to extract the edge of MRI image as well as the image of Lena. The experimental results indicate the algorithm can considerably improve the edge resolution of the traditional morphological edge detection methods and is practical.     

A face recognition method based on modular PCA and SVD

In order to solve the problem that modular PCA method is sensitive to translation, rotation and other geometric transform, a face recognition method based on modular PCA and singular value decomposition (SVD) is proposed. The PCA features of sub image and SVD features are extracted respectively. The distance measure that fuses information of modular PCA and SVD is obtained. Minimum distance classifier is used to face recognition. Experimental results on ORL human face database show that the proposed method can obtain higher recognition rate.     

Single sample face recognition based on B-spline and image gradient

A single sample face recognition algorithm based on B-spline and image gradient is proposed. Image gradient method for face recognition has advantage of illumination invariant. But the recognition rate will be greatly decreased when the image contains noise which will seriously influence gradient information. Traditional methods to reduce noise smooth image at the same time and image gradient reorganization rate will be reduced. As the B-spline filter has the feature which can adjust the order, B-spline filter with different orders can be selected according to the image noise value to minimize noise while preserve image gradient information. Experiments prove that using B-spline and image gradient algorithm can achieve a better recognition rate than traditional filtering method on single sample face recognition problem.     

Automatic head recognition by integrating mean shift with multiple features

To improve the head detection accuracy in video sequences captured with fixed vertical monocular camera, a novel method of head recognition based on mean shift and multiple features is proposed. Firstly, mean shift based image segmentation algorithm with color information and spatial information is suggested to derive the candidate head components in images. Furthermore, by referring to two features that the contour of human head regions are approximate round and the hair color distribution is clustered, the evaluation models based on the contour information and inside color information of candidate head components are presented for head recognition. The experimental results show that the proposed algorithm can effectively reduce the light interfere and eliminate fake target whose color information is similar to hair color distribution. The detection rate for static images can reach about 89.4%.     

基于几何语义知识的冬小麦自动分类

【研究目的】为了实现遥感影像的作物自动分类,并探索空间信息在分类中作用,【方法】本文提出结合光谱和空间信息的作物分类方法。首先,借助光谱信息实现地物初始分割,然后以目标作物历史空间分布为语义约束,根据隶属度提取目标作物。最后,在多时相遥感影像条件下,以冬小麦为目标作物进行了方法的验证,【结果】结果显示,本文方法可实现冬小麦自动提取与识别,总体精度为95.33%,Kappa系数为0.90,可满足农情监测的实际需求。另外,在单时相遥感影像条件下,本文结合几何语义知识的作物分类精度也达到了较高水平。【结论】相对于遥感影像单一光谱信息的分类方法,本文方法利用了作物空间信息,不仅能满足精度要求,还实现了分类的自动化,对工程化应用具有一定的参考价值。     

Human computer interaction with face poses estimation

Focusing on the problem of computer operation without hand for the disabled people, a new method of mouse control based on visual analysis of the face poses is proposed. First, the face images from CCD camera are obtained and the face areas are located based on AdaBoost algorithms with Haar features. Then, 5 feature points including 2 eye corners, 2 mouth corners and nose tip are detected with image analysis. After that, with help of the feature points in a predefined frontal face image, the face poses are estimated in the real-time face images. Finally, the 3 deflection angles are adopted to define human computer interactive information, which can be used to determine positions and operations of the mouse in the human computer interactive interface. The experimental results demonstrate the efficiency of our method that the face poses can be estimated accurately as well as the operation information to the mouse be achieved rapidly. It is valuable in human computer interaction for the disabled people.     

A method for feature contour extraction of triangular meshesbased on multi seed points

An algorithm for extracting feature curves of triangular meshes based on multi seed points is proposed. First, feature contour positions of triangular meshes are detected on the basis of curvature value and normal vector, and these positions are highlighted with difference colours. Then, several seed points on the proper position are assigned and their feature attributes are computed. Next, their node assessment functions are constructed and correct path nodes are chosen. Finally, the curves extracted from the triangular meshes are smoothed. Experimental results show that the method is not only effective to extract closed feature curves and sharp edges, but also useful for opened feature curves and the region that the features change gently. It has higher accuracy than other methods to detect feature at the branch feature regions.     

Fusion method for infrared and visible light images based on image segmentation and stationary wavelet transform

A novel image fusion method based on image segmentation and stationary wavelet transform (SWT) is proposed to improve the visual effect of fused infrared and visible light images. Infrared image is firstly separated into object and background region utilizing Otsu combined with edge detection. Then a multiresolution decomposition using SWT is made to the background region of the infrared image and the visible light image. Neighborhood spatial frequency and absolute value are adopted as fusion rules in low-frequency and high-frequency coefficients. The background fused image is reconstructed by inverse SWT. The final infrared and visible light fused image is obtained by fusing the background fused image and the object region of infrared image base on weighted fusion rule. The experimental results show that the object information of the infrared image is obviously highlighted and the scene information of the visible light image is well represented. The visual effect of fused image is improved efficiently by utilizing the proposed method. The proposed method works better than the traditional Laplacian Pyramid and wavelet transform fusion algorithms in terms of standard deviation, comentropy and mutual information. Experimental results verify its effectiveness.     

基于图像特征的收获前籽棉品级聚类融合分析

为客观评价收获前籽棉品级,依据籽棉品级国家标准,基于机器视觉在3个颜色空间中选取棉棉瓣大小、色泽特征用K-均值、竞争学习网络方法对7个品级的样本进行聚类融合分析。结果表明,亮度修正后特征之间极显著相关,Hunter颜色空间较好。肉眼对第1、2、7品级的识别率为73%~100%,3~6品级为26%~46%,总计47.7%;聚类融合对各品级的识别率为65%~100%,总计78.6%。聚类融合方法基于人类的先验知识,在更宽的视觉范围内更均衡所有特征,可克服个体聚类器的过度训练,能够客观地识别收获前籽棉品级,提高其采摘、收购质量。     

基于数字图像的玉米叶面积测量方法研究

【研究目的】以数字图像处理技术为基础,采用数码相机获取数字图像,建立使用Visual C++6.0开发的图像处理软件进行玉米叶面积测量的方法。【方法】首先,用数码相机拍摄挂在墙上粘贴着玉米叶片的参照白板来获得数字图像。然后提取图像的蓝色分量,利用中值滤波、膨胀、腐蚀等方法去除图像噪声,最大类间方差法自动阈值分割目标。接着进行区域标记与区域像素数统计,区域像素数结合图像分辨率就可以获得玉米叶片的实际面积。【结果】与直尺法所测结果具有很好的相关性,且比直尺法更精确,【结论】此种拍摄方法避免了几何畸变的产生。蓝色分量图像中目标和背景的灰度差异最大,便于分割。     

无人机影像拼接软件在农业中应用的比较研究

无人机遥感为精准农业管理提供了新的工具。实现无人机影像高精度自动拼接是开展无人机遥感应用的基础。目前,已有不同无人机影像拼接软件在几何精度斱面性能比较的研究,但关于光谱精度斱面还未有相关研究,而其对定量遥感反演非常重要。本研究比较了目前最受欢迎的两款无人机拼接软件Pix4D和Photoscan在农业应用时,拼接影像过程对原始影像光谱信息的影响,以为用户推荐最优软件。为此,基于冬小麦氮肥梯度试验,本研究在小麦生长季利用无人机搭载多光谱传感器获取相关影像,幵将其分别基于Pix4D软件, Photoscan软件拼接处理。同时,基于传感器厂商提供的单张影像处理技术,将每次传感器拍摄数据处理成未拼接的单张多光谱影像。基于以上数据,在施肥处理小区随机布设样点,分别提取3种类型影像上的样点光谱信息,比较它们光谱反射率及其对比度的差异。结果表明,尽管Pix4D拼接影像和Photoscan拼接影像各波段光谱反射率都与单张影像的反射率有显著相关性,但与Pix4D拼接影像相比, Photoscan拼接影像的光谱反射率和变异系数与原始单张影像之间更为接近。Photoscan能保留更多的原始光谱信息。结合已有关于两款软件在几何精度和价格斱面的比较研究,本研究推荐Photoscan为农业应用时的最优性价比软件。     

A Vector Quantification Image Coding Scheme Based on Wavelet Transform

The paper proposes an image compression and coding scheme based on discrete wavelet transform (DWT), which makes the best of the relativities among original image and coefficients of wavelet, the relativities among pixels of subimages, the relativities among directions of subimages. Human visual specialities were considered, so the compression scheme leads up to make full use of the statistical redundancy and visual redundancy of image. To get high compression ratio and good quality of image, the coefficients of image working on the quality of image greatly were reserved accurately, but the coefficients of image working on the quality of image slightly were quantified roughly. Also, the regrouping and coding of the coefficients could be used universally. The experimental results show that the compression method is provided with simply calculating, little time for coding and decoding and satisfying quality companied with high compression ratio.     

不同棉花群体冠层数字图像颜色变化特征研究

 为了探索利用数字图像监测棉花长势的方法,研究了不同群体的冠层数字图像颜色变化特征。结果表明,RGB模型中,R分量在全生育期呈现开口向上的二次曲线变化趋势;G分量变化为开口向下的二次曲线;不同密度B分量差异较大,变化曲线形状介于R和G之间。RGB组合性状中,R/(G+B)、R/G、R、R/B类似,B、B/(G+R)、B/G类似,G、G/(R+B)类似,其中R/G、G/(R+B)、R/(G+B)的二次曲线特征更明显。HIS模型中,H分量变化趋势为开口向上的二次曲线;I分量变化基本符合二次曲线;S分量变化无二次曲线特征。R,G,R/G、G/(R+B)、R/(G+B)以及H分量在整个生育期变化具有一定的规律可循,可能是反映作物长势的潜在指标,这将对数字图像监测作物长势具有重要意义。     

Lossless data hiding with ranks transform

For taking full advantage of carrier image's pixel correlation, a lossless data hiding algorithm with ranks transform (LDHA_RT) is proposed. Along with three directions of column, row and diagonal, the pixel difference images and their histograms are calculated one after another. The obtained pixel difference histograms are shifted to construct the redundancy spaces, so that the hiding information could be embedded in a designated position. After the hiding data is extracted, the carrier image could be recovered reversibly without any additional information. The experiments demonstrate that the PSNR/bits of the LDHA_RT is the lowest or close to the lowest, compard with these algorithms mentioned in the references. This implies that the LDHA_RT has large capacity of hiding data under a good image quality.     

Optimization of regularization parameters for electricalimpedance tomography

The image quality and computation speed are bounded up with regularization parameters. To improve the ill-posed property of the inverse problem of electrical impedance tomography (EIT), a novel approach, which is based on the product of the residual norm and the solution norm(PRS), is presented to optimize the Tikhonov regularization parameters of EIT. To verify the feasibility and effectiveness of the proposed method, five simulations of image reconstruction, together with a tank experiment, have been carried out with considering different sizes, locations, conductivity distributions and numbers of the target areas as well as the scenarios of the data with noises. The encouraging results demonstrate that the proposed optimization approach can identify the relatively optimal regularization parameter quickly and has better noise immunity, and it also enhances the quality of the reconstructed images significantly compared with the conventional L-curve method.