An adaptive median filter algorithm based on B-spline function

According to the B-spline convolution mask, first, the contrast sensitiveness (CS) is computed and then is viewed as a noise sensitiveness coeficient (NSC) to adaptively determine a noise-recognized threshold value. Based on the noise density function (NDF) in a 3×3 window, the filtering window size is adaptively adjusted, and then a median filter is used to eliminate the noise-marked pixels. The experiment results show that the proposed algorithm can preserve image detail information well and effectively remove the noises, particularly the impulse noises that is also called salt-and-pepper noises superimposed on the computed tomography (CT) and magnetic resonance imaging (MRI) medical images.     

Study on the distribution law of random code structure of irregular LDPC codes and its application in eliminating short cycles

The distribution law of the random code structure of randomly constructed irregu- lar low-density parity-check (LDPC) codes is studied. Based on the Progressive Edge-Growth (PEG) algorithm, a new algorithm which can both eliminate short cy- cles and keep the distribution of the random code structure is presented. The ex- perimental results show that the performance of the irregular LDPC codes con- structed by the new algorithm is superior to that of the PEG algorithm.     

A coarse-to-fine image registration method based on visual attention model

Image registration is fundamental and crucial to remote sensing. However getting highly accurate registration performance automatically and fast for large-field images consistently is a challenge. As a work around to this problem, we propose a new image registration concept based on visual attention in this paper. This concept employs the advantages of feature-based or area-based methods to improve the precision and efficiency of image registration. The key concept of proposed integrated scheme is to make optimum use of the highly prominent details in the full scene by means of visual attention computational mechanism. To testify the validation, comparisons with other classical methods are carried out on real-world images. The experimental results show that the proposed method can effectively perform on multi-view/multi-temporal remote sensing images with outstanding precision and time saving performance.     

Sparse representation based on projection method in online least squares support vector machines

A sparse approximation algorithm based on projection is presented in this paper in order to overcome the limitation of the non-sparsity of least squares support vector machines （LS-SVM）. The new inputs are projected into the subspace spanned by previous basis vectors （BV） and those inputs whose squared distance from the subspace is higher than a threshold are added in the BV set, while others are rejected. This consequently results in the sparse approximation. In addition, a recursive approach to deleting an exiting vector in the BV set is proposed. Then the online LS-SVM, sparse approximation and BV removal are combined to produce the sparse online LS-SVM algorithm that can control the size of memory irrespective of the processed data size. The suggested algorithm is applied in the online modeling of a pH neutralizing process and the isomerization plant of a refinery, respectively. The detailed comparison of computing time and precision is also given between the suggested algorithm and the nonsparse one. The results show that the proposed algorithm greatly improves the sparsity just with little cost of precision.     

A low-complexity sensor fusion algorithm based on a fiber-optic gyroscope aided camera pose estimation system

Visual tracking, as a popular computer vision technique, has a wide range of applications, such as camera pose estimation. Conventional methods for it are mostly based on vision only, which are complex for image processing due to the use of only one sensor. This paper proposes a novel sensor fusion algorithm fusing the data from the camera and the fiber-optic gyroscope. In this system, the camera acquires images and detects the object directly at the beginning of each tracking stage; while the relative motion between the camera and the object measured by the fiber-optic gyroscope can track the object coordinate so that it can improve the effectiveness of visual tracking. Therefore, the sensor fusion algorithm presented based on the tracking system can overcome the drawbacks of the two sensors and take advantage of the sensor fusion to track the object accurately. In addition, the computational complexity of our proposed algorithm is obviously lower compared with the existing approaches(86% reducing for a 0.5 min visual tracking). Experiment results show that this visual tracking system reduces the tracking error by 6.15% comparing with the conventional vision-only tracking scheme(edge detection), and our proposed sensor fusion algorithm can achieve a long-term tracking with the help of bias drift suppression calibration.     

Tire Defect Detection Using Local and Global Features

In this paper, we present a tire defect detection algorithm based on sparse representation. The dictionary learned from reference images can efficiently represent the test image. As the representation coefficients of normal images have a specific distribution, the local feature can be estimate by comparing representation coefficient distribution. Meanwhile, a coding length is used to measure the global features of representation coefficients. The tire defect is located by both these local and global features. Experimental results demonstrate that the proposed method can accurately detect and locate the tire defects.     

Anisotropic Diffusion Filtering for Reconstruction of Poisson Noisy Sinograms

Anisotropic diffusion filtering is a powerful tool for enhancing image while preserving edges not drifting or blurred. In the paper the tool is applied in rvconstruction of Poisson noisy sinograms which have not too many projections. The proposed algorithm has three steps： extending sinogram with interpolating operation, enhancing sinogram by anisotropic diffusion filtering and rvconstructing by filtered background projection （FBP）. It shows that the algorithm can get rvconstruction images which have much better in quality than the algorithm which only uses FBP. The algorithm is saliently morv efficient than statistical tomography algorithms as well. We illustrate the performance of the algorithm on the data generated by differvnt moralities.     

A Constraint Satisfaction Neural Network and Heuristic Combined Approach for Concurrent Activities Scheduling

Scheduling activities in concurrent product development process is of great significance to shorten development lead time and minimize the cost. Moreover, it can eliminate the unnecessary redesign periods and guarantee that serial activities can be executed as concurrently as possible. This paper presents a constraint satisfaction neural network and heuristic combined approach for concurrent activities scheduling. In the combined approach, the neural network is used to obtain a feasible starting time of all the activities based on sequence constraints, the heuristic algorithm is used to obtain a feasible solution of the scheduling problem based on resource constraints. The feasible scheduling solution is obtained by a gradient optimization function. Simulations have shown that the proposed combined approach is efficient and feasible with respect to concurrent activities scheduling.     

Grasping Force Optimization Algorithm of Soft Multi-fingered Hand Based on Safety Margin Detection

The classical gradient flow optimization algorithm requires a valid initial point before starting the recursive algorithm,and the existing methods can’t guarantee that the initial values fully satisfy the friction cone constraints of contact point in the optimization process of gradient flow algorithm.In order to improve safety margin and prevent the finger from slipping at contact point,we present an iterative method of safe initial values with safety margin detection and develop a gradient flow optimization algorithm based on the safe initial values.Firstly,the safety margin is defined which more intuitively reflects the margin of the grasping forces at contact point.The resulting safe initial values can be achieved by the detection of desired safety margin at each iteration.Secondly,the safe initial values are usually not optimal,even with the valid initial values,and it can’t always ensure that the finger contact force always satisfies the friction cone constraints during the optimization.It is an effective way to eliminate the unreliable initial values in the optimization and obtain a safer initial values by increasing the safety margin.By transforming the safe initial values into an initial point of the gradient flow algorithm,the final optimized values of grasping forces can be generated efficiently by gradient flow iteration.Grasp examples of the soft multi-fingered hand indicate the effectiveness of the general solution of the force optimization algorithm based on safety margin detection.The method eliminates the shortcomings of the gradient flow optimization process caused by the initial value problem and provides a more accurate and reliable force optimization result for multi-fingered dexterous manipulation.     

Visual salience guided feature-aware shape simplification

In the area of 3D digital engineering and 3D digital geometry processing, shape simplification is an important task to reduce their requirement of large memory and high time complexity. By incorporating the content-aware visual salience measure of a polygonal mesh into simplification operation, a novel feature-aware shape simplification approach is presented in this paper. Owing to the robust extraction of relief heights on 3D highly detailed meshes, our visual salience measure is defined by a center-surround operator on Gaussian-weighted relief heights in a scale-dependent manner. Guided by our visual salience map, the feature-aware shape simplification algorithm can be performed by weighting the high-dimensional feature space quadric error metric of vertex pair contractions with the weight map derived from our visual salience map. The weighted quadric error metric is calculated in a six-dimensional feature space by combining the position and normal information of mesh vertices. Experimental results demonstrate that our visual salience guided shape simplification scheme can adaptively and effectively re-sample the underlying models in a feature-aware manner, which can account for the visually salient features of the complex shapes and thus yield better visual fidelity.     

彩色图像去雾清晰化研究

为了解决雾霾天气的影响下成像设备采集的图像容易出现的降质及色彩失真问题, 并从有雾图像中增强其细节信息, 提高原图像的对比度和清晰度. 将彩色图像RGB通道分别做相应的图像增强算法处理, 全局直方图均衡化把整体的灰度直方图进行均匀分布的处理, 小波变换算法对彩色图像进行多层次分解, 多尺度Retinex算法通过高斯函数做卷积运算对图像做多尺度变换. 实验结果表明, 全局直方图均衡化、小波变换算法和多尺度Retinex算法都能增强雾天图像的景物信息, 有对应于各自的优势和不足. 相比较这3种算法而言, 多尺度Retinex算法得到的去雾图像亮度增强、细节信息突出、失真度小, 能更好地进行去雾增强.     

基于深度学习的遥感图像去雾算法

针对遥感图像由于雾霾的存在导致图像清晰度下降的问题,提出了一种基于深度学习的图像去雾算法.首先将原有大气散射模型进行变形得到一个端到端的去雾模型,再将多个未知参数统一在一个参数中,运用多尺度卷积神经网络对未知参数进行估计,最后将参数估计值代入去雾模型中得到无雾图像.针对无参考图像数据集,先运用现有数据集对网络进行初步训练,再加入自建数据集对网络进行二次训练.实验结果表明,与相关去雾算法进行对比,该算法在视觉效果和客观指标上都有不同程度的提高,有效提升了遥感图像在雾霾天气状况下的清晰度.     

MSIMCNN: Multi-scale inception module convolutional neural network for multi-focus image fusion

The aim of image fusion is to obtain a clear image by combining useful information coming from multiple images. However, the fused image usually has the problem of artifacts and unclear boundary. To address these problems, a deep convolutional neural network based framework for multi-focus image fusion is proposed in this paper, called multi-scale inception module convolutional neural network (MSIMCNN). MSIMCNN converts the entire image into a binary mask to estimate the focus characteristics, and obtains the clear boundary between focus and defocus. First of all, a pair of focus images and the corresponding feature images detected by the Laplace operator are inputted into the network. The Laplace operator can detect the edge and gradient of focus in the image, which can help us accurately reconstruct the focused area in the focus map and distinguish the focus and defocus boundaries. Then, in the feature extraction stage, different scales of convolution kernels are designed to extract the rich and complementary features at different scales of the source images. At the same time, the inception module is added to increase the width of the network and reduce the parameters, which can extract more focus features required for image reconstruction and reduce the complexity. Finally, the focus map of the source image pair can be obtained in the feature reconstruction stage. In this stage, an efficient method is proposed to make the focus mask, which is used for the calculation of the loss function and the generation of the training set. The experimental results on different data sets confirm the superiority and effectiveness of MSIMCNN compared with other methods.

     

多尺度密集网络在红外和可见光图像融合应用

为了进一步提升红外和可见光图像的融合效果,提出了一种基于多尺度卷积算子和密集连接网络的图像融合模型.该模型首先使用多尺度卷积算子计算图像的直接多尺度特征,然后使用密集连接网络计算图像的间接多尺度特征.为了得到图像像素信息在不同尺度下的融合权重,通过叠加的方式将各个尺度密集连接网络的输出进行融合,并使用活动图方法计算两类图像的融合权重,最后根据权重计算结果得到融合图像,实验在THO数据集和CMA数据集获得较好的识别率.     

基于泰勒展开式与曲率逆向驱动的图像超分辨算法

针对传统的插值算法和基于模型的超分辨算法通常会导致图像对比度和清晰度下降等问题,提出了一种基于泰勒展开式与曲率逆向驱动图像超分辨算法。该算法首先采用泰勒公式估计图像灰度值的变化趋势,然后利用等照度线的曲率细化图像边缘特征,并将梯度作为约束条件抑制图像边缘的锯齿与振铃效应。大量实验表明,所提算法在清晰度和信息保留度上比传统算法更具有优势,算法处理结果更符合人眼视觉效果,在泰勒展开式的基础上进行逆向扩散也使该算法的运行效率明显高于传统迭代算法。     

基于局部加权拟合算法的无人机遥感影像多尺度检测技术

目前提出的无人机遥感影像多尺度检测技术平均图像灰度较差,导致检测结果清晰度较低;为了解决上述问题,基于局部加权拟合算法研究了一种新的无人机遥感影像多尺度检测技术,选用最小二乘法进行多次循环计算,确定周围区域重复率,通过抽稀处理提高数据精度;根据高斯金字塔得到n阶的影像序列,利用高斯金字塔和差分尺度划分完成遥感影像的特征提取;引入加权拟合算法,构建有效影像数据集,确定影像网络模型,从而完成合并,实现影像数据的检测;实验结果表明,基于局部加权拟合算法的无人机遥感影像多尺度检测技术能够有效提高平均图像清晰度,增强检测结果的清晰度。     

基于多尺度自商图和改进积分投影法的人眼定位

针对传统的投影方法在人眼定位时易受光照干扰以及难以获得精准的人眼中心点的问题,提出一种基于多尺度自商图和改进的积分投影法的人眼定位算法.首先利用多尺度自商图消除人脸图像的光照影响;然后分析眼睛在水平方向上灰度分布的特点,采用两个行梯度算子对积分投影法进行了改进,以提升眼睛区域特征并初步定位人眼区域;接着采用Sobel算子对人眼区域进行滤波得到人眼滤波图,并对人眼滤波图的垂直积分投影曲线进行高斯函数拟合,根据拟合结果分割出左眼窗口和右眼窗口;最后,计算左眼窗口和右眼窗口的尺寸,获取左眼窗口和右眼窗口的中心点,即为人眼中心点.在YaleB人脸数据库和JAFFE人脸数据库上测试表明,本文方法对复杂光照、人脸边缘以及人脸表情适应性强,可以获得较为精准的人眼中心点.     

一种山区图像边缘信息的分水岭变换提取方法

为了能够抑制引起过分割的根本原因-噪声和细密纹理,提出了一种改进的提取山区图的山脊线、山谷线的边缘信息的分水岭算法,该算法着重于解决传统分水岭的过分割问题;该算法首先对梯度图进行一种改进的多尺度梯度算子处理,以减小图像噪声对分水岭的过度分割的影响;然后利用基于标记符控制的最小强制技术去除由噪声和不必要的图像细节形成的谷底,使修正后的梯度图像只对应包含区域极大值的标记图像,改进后的方法能准确地提取出所需要的关键边缘信息.     

多尺度双通道卷积神经网络下的刺绣模拟

目的 针对现有刺绣模拟算法中针线感不强、针线轨迹方向单一等问题,提出了一种基于多尺度双通道卷积神经网络的刺绣模拟算法。方法 1）搭建多尺度双通道网络,选取一幅刺绣艺术作品作为风格图像,将MSCOCO（microsoft common objects in context）数据集作为训练集,输入网络得到VGG（visual geometry group）网络损失和拉普拉斯损失;2）将总损失值传回到网络,通过梯度下降法更新网络参数,并且重复更新参数直到指定的训练次数完成网络训练;3）选取一幅目标图像作为刺绣模拟的内容图像,输入训练完成的网络,获得具有刺绣艺术风格的结果图像;4）使用掩模图像将得到的结果图像与绣布图像进行图像融合,即完成目标图像的刺绣模拟。结果 本文算法能产生明显的针线感和多方向的针线轨迹,增强了刺绣模拟绘制艺术作品的表现力。结论 本文将输入图像经过多尺度双通道卷积神经网络进行前向传播,并使用VGG19、VGG16和拉普拉斯模块作为损失网络进行刺绣模拟。实验结果表明,与现有卷积神经网络风格模拟算法对比,本文提出的网络能够学习到刺绣艺术风格图像的针线特征,得到的图像贴近真实刺绣艺术作品。     

基于生成式对抗网络的道路交通模糊图像增强

为了提高道路交通模糊图像增强的质量,进一步促进道路交通管理,针对道路交通场景下的运动模糊图像增强问题,提出了一种基于生成式对抗网络的多尺度多路径学习的模型。首先,选用具有多尺度卷积核的神经网络,对输入的图像进行更细致地特征值提取;其次,将局部残差学习和全局残差学习相结合,采用多路径多权重共享的递归学习,并利用判别网络和生成网络间的对抗训练优化网络参数;最后,实现端到端直接生成图像。实验结果表明:提出的模型可以有效地增强道路交通场景下的运动模糊图像,生成的图像细节更加丰富,具有较好的图像视觉效果。