Multiobjective neural network for image reconstruction

The authors propose a multiobjective neural network model and algorithm for image reconstruction from projections. This model combines the Hopfield model and multiobjective decision making approach. A weighted sum optimisation based neural network algorithm is developed. The dynamic process of the net is based on minimisation of a weighted sum energy function and Euler's iteration and this algorithm is applied to image reconstruction from computer-generated noisy projections and Siemens Somaton DR scanner data, respectively. Reconstructions based on this method are shown to be superior to those based on conventional iterative reconstruction algorithms such as MART (multiplicate algebraic reconstruction technique) and convolution from the point of view of accuracy of reconstruction. Computer simulation using the multiobjective method shows a significant improvement in image quality and convergence behaviour over conventional algorithms     

基于并行BP神经网络的近红外光断层图像重建方法基础研究

提出一种基于并行BP神经网络的近红外光断层成像(Near-infrared optical tomography,NIR OT)图像重建算法,利用BP神经网络来表征生物组织内部光学参数的空间分布和边界光强之间的非线性映射关系.该方法将一个复杂的模型分解成简单的模型分别建立并行的神经网络.利用Femlab软件完成基于有限元的稳态扩散方程的两个简单模型的正向问题求解,根据提出的平均优化散射系数和正向问题训练的大量数据集合,建立并训练并行神经网络,通过对两个网络结果的分析,实现快速获得更复杂模型的光学参数的重构.算法能够快速识别特异组织的位置和准确反映热疗过程中生物组织的优化散射系数的变化趋势.     

A neural network image reconstruction technique for electrical impedance tomography

Reconstruction of images in electrical impedance tomography requires the solution of a nonlinear inverse problem on noisy data. This problem is typically ill-conditioned and requires either simplifying assumptions or regularization based on a priori knowledge. The authors present a reconstruction algorithm using neural network techniques which calculates a linear approximation of the inverse problem directly from finite element simulations of the forward problem. This inverse is adapted to the geometry of the medium and the signal-to-noise ratio (SNR) used during network training. Results show good conductivity reconstruction where measurement SNR is similar to the training conditions. The advantages of this method are its conceptual simplicity and ease of implementation, and the ability to control the compromise between the noise performance and resolution of the image reconstruction.     

一种改进的神经网络相关性剪枝算法

神经网络的剪枝有利于网络结构的简化,而作为剪枝算法中的比较重要的相关性剪枝算法,在计算了隐层节点输出的线性相关性和方差后,对于如何根据线性相关值和方差值删除节点并没有给出明确的界限。文章通过研究神经网络的相关性剪枝算法,给出一种以网络的误差传递为思想,根据方差值删除节点的方法,并通过实验证明,该方法不仅能够有效的简化网络结构,保证网络精度,而且计算简单。     

An artificial neural network model for effective dielectricconstant of microstrip line

A backpropagation network structure is presented for the calculation of the effective dielectric constant (ϵ_eff) of microstrip lines. Results of the network are compared with those of the spectral-domain (SD) technique     

Saliency-directed color image interpolation using artificial neural network and particle swarm optimization

In this study, a saliency-directed color image interpolation approach using artificial neural network (ANN) and particle swarm optimization (PSO) is proposed. First, a high-quality saliency map of a color image to be interpolated is generated by a modified block-based visual attention model in an effective manner. Then, based on the saliency map, bilinear interpolation and ANN-PSO interpolation are employed for non-saliency (non-ROI) and saliency (ROI) blocks, respectively, to obtain the final color interpolation results. In the proposed ANN-PSO interpolation scheme, ANN is used to determine the orientation of each 5 × 5 image pattern (block), whereas PSO is employed to determine the weights in 5 × 5 interpolation filtering masks. The proposed approach is applicable to image interpolation with arbitrary magnification factors (MFs). Based on the experimental results obtained in this study, the color interpolation results by the proposed approach are better than those by five comparison approaches.     

基于人工神经网络的图像检索平台的设计与开发

用户描述图像的高层抽象语义与图像内在的底层特征之间存在差异,此时仅依靠图像内容特征进行检索的系统无法准确完成用户的检索任务。针对以上问题,提出了使用神经网络进行图像的匹配计算方法,通过样例自动学习和用户反馈学习两种学习方式,形成图像底层特征到图像分类的正确映射,学习后的神经网络可以进行图像的自动分类及检索。该方法结合了图像的底层特征描述及用户的高层语义反馈,有效地弥补了语义鸿沟。最后,系统通过整合Web前端、图像提取模块、神经网络模块及数据库模块,实现了神经网络学习及图像检索的完整流程。     

人工神经网络概述

近年来,随着人工神经网络在信息、自动化、医学、经济等领域的广泛应用和杰出表现,使得人工神经网络又开始得到广泛关注和重新重视。基于需求,本文对人工神经网络做了简单概述。     

BFA based neural network for image compression

A novel Bacterial Foraging Algorithm （BFA） based neural network is presented for image compression. To improve the quality of the decompressed images, the concepts of reproduction, elimination and dispersal in BFA are firstly introduced into neural network in the proposed algorithm. Extensive experiments are conducted on standard testing images and the results show that the proposed method can improve the quality of the reconstructed images significantly.     

Fast image reconstruction network in image stitching

Compared with the traditional feature-based image stitching algorithm, the free-view image stitching algorithm based on deep learning has the advantages of fast stitching speed and good effect. However, these algorithms still cannot achieve real-time splicing speed. For the image reconstruction stage, we redesign a new fast image reconstruction network. This network is designed based on ShuffleNet, and the new network structure and loss function will reduce the time required for image reconstruction. In addition, this network can also reduce the performance loss after the network is lightweight. It is proved by experiments that the fast image reconstruction network can realize real-time high-resolution free-view image reconstruction.     

An evaluation of maximum likelihood reconstruction for SPECT

A reconstruction method for SPECT (single photon emission computerized tomography) that uses the maximum likelihood (ML) criterion and an iterative expectation-maximization (EM) algorithm solution is examined. The method is based on a model that incorporates the physical effects of photon statistics, nonuniform photon attenuation, and a camera-dependent point-spread response function. Reconstructions from simulation experiments are presented which illustrate the ability of the ML algorithm to correct for attenuation and point-spread. Standard filtered backprojection method reconstructions, using experimental and simulated data, are included for reference. Three studies were designed to focus on the effects of noise and point-spread, on the effect of nonuniform attenuation, and on the combined effects of all three. The last study uses a chest phantom and simulates Tl-201 imaging of the myocardium. A quantitative analysis of the reconstructed images is used to support the conclusion that the ML algorithm produces reconstructions that exhibit improved signal-to-noise ratios, improved image resolution, and image quantifiability.     

An image analyzing system using an artificial neural network forevaluating the parenchymal echo pattern of cirrhotic liver and chronichepatitis

To objectively evaluate the parenchymal echo pattern of cirrhotic liver and chronic hepatitis, the authors applied an image analyzing system (IAS) using a neural network. Autopsy specimens in a water tank (n=13) were used to examine the relationship between the diameter of the regenerative nodule and the coarse score (CS) calculated by IAS. CS was significantly correlated with the diameter of the regenerative nodule (p<0.0001, r=0.966). CS is considered to be useful for evaluating the coarseness of the parenchymal echo pattern     

TBNN: totally-binary neural network for image classification

Most binary networks apply full precision convolution at the first layer. Changing the first layer to the binary convolution will result in a significant loss of accuracy. In this paper, we propose a new approach to solve this problem by widening the data channel to reduce the information loss of the first convolutional input through the sign function. In addition, widening the channel increases the computation of the first convolution layer, and the problem is solved by using group convolution. The experimental results show that the accuracy of applying this paper''s method to state-of-the-art (SOTA) binarization method is significantly improved, proving that this paper''s method is effective and feasible.     

基于门控卷积神经网络的图像超分辨重建算法北大核心CSCD

近年来,卷积神经网络被广泛应用于图像超分辨率领域。针对基于卷积神经网络的超分辨率算法存在图像特征提取不充分,参数量大和训练难度大等问题,本文提出了一种基于门控卷积神经网络(gated convolutional neural network, GCNN)的轻量级图像超分辨率重建算法。首先,通过卷积操作对原始低分辨率图像进行浅层特征提取。之后,通过门控残差块(gated residual block, GRB)和长短残差连接充分提取图像特征,其高效的结构也能加速网络训练过程。GRB中的门控单元(gated unit, GU)使用区域自注意力机制提取输入特征图中的每个特征点权值,紧接着将门控权值与输入特征逐元素相乘作为GU输出。最后,使用亚像素卷积和卷积模块重建出高分辨率图像。在Set14、BSD100、Urban100和Manga109数据集上进行实验,并和经典方法进行对比,本文算法有更高的峰值信噪比(peak signal-to-noise ratio,PSNR)和结构相似性(structural similarity,SSIM),重建出的图像有更清晰的轮廓边缘和细节信息。     

A NEURAL NETWORK HARDWARE MODEL FOR IMAGE THINNING

A new parallel thinning algorithm was proposed.This paper suggests a set of 5x5cloning templates of cellular neural network for image thinning.The principle of the templatesdesign and hardware model are proposed in the paper.Although the connected weights betweenthe neurons are asymmetric,it is shown that the network is stable,and it can be easily realized.     

Detail retaining convolutional neural network for image denoising

Compared with the traditional image denoising method, although the convolutional neural network (CNN) has better denoising performance, there is an important issue that has not been well resolved: the residual image obtained by learning the difference between noisy image and clean image pairs contains abundant image detail information, resulting in the serious loss of detail in the denoised image. In this paper, in order to relearn the lost image detail information, a mathematical model is deducted from a minimization problem and an end-to-end detail retaining CNN (DRCNN) is proposed. Unlike most denoising methods based on CNN, DRCNN is not only focus to image denoising, but also the integrity of high frequency image content. DRCNN needs less parameters and storage space, therefore it has better generalization ability. Moreover, DRCNN can also adapt to different image restoration tasks such as blind image denoising, single image superresolution (SISR), blind deburring and image inpainting. Extensive experiments show that DRCNN has a better effect than some classic and novel methods.     

人工神经网络法LCD色度特性化研究

目前国内外LCD的特性化研究方法已经很多.本文采用人工神经网络法定量描述了液晶显示器的颜色特性.对比S-shape模型,在实验测量的基础上,比较了两种方法的优缺点.神经网络主要研究了训练方法、传递函数、隐层数和隐层单元数对结果的影响,最后得到人工神经网络法的颜色校正模型.     

System calibration and statistical image reconstruction for ultra-high resolution stationary pinhole SPECT

For multipinhole single-photon emission computed tomography (SPECT), iterative reconstruction algorithms are preferred over analytical methods, because of the often complex multipinhole geometries and the ability of iterative algorithms to compensate for effects like spatially variant sensitivity and resolution. Ideally, such compensation methods are based on accurate knowledge of the position-dependent point spread functions (PSFs) specifying the response of the detectors to a point source at every position in the instrument. This paper describes a method for model-based generation of complete PSF lookup tables from a limited number of point-source measurements for stationary SPECT systems and its application to a submillimeter resolution stationary small-animal SPECT system containing 75 pinholes (U-SPECT-I). The method is based on the generalization over the entire object to be reconstructed, of a small number of properties of point-source responses which are obtained at a limited number of measurement positions. The full shape of measured point-source responses can almost be preserved in newly created PSF tables. We show that these PSFs can be used to obtain high-resolution SPECT reconstructions: the reconstructed resolutions judged by rod visibility in a micro-Derenzo phantom are 0.45 mm with 0.6-mm pinholes and below 0.35 mm with 0.3-mm pinholes. In addition, we show that different approximations, such as truncating the PSF kernel, with significant reduction of reconstruction time, can still lead to acceptable reconstructions.      

基于神经网络的图像识别

神经网络图像识别技术是随着当代计算机技术、图像处理、人工智能、模式识别理论等发展起来的一种新型图像识别技术。在进行图像识别之前需要利用数字图像处理技术进行图像预处理以及特征提取。本文选取字符图像0~9作为识别目标,对图像预处理过程进行了叙述,并在此基础上选取字符图像矩阵每行的与每列的黑色像素点之和以及图像欧拉数这两个特征作为BP神经网络的输入样本。经实验仿真表明图像的平均识别率为89%,这表明图像预处理的结果和提取的特征是合适的、有效的,设计的BP网络也较好的完成了模式分类识别工作。