基于散斑图像梯度域的数字图像相关盲去模糊方法EI北大核心CSCD

针对复杂测量环境或高动态测量过程中出现的运动模糊问题,提出了一种灰度稀疏先验与参考图像梯度域先验相结合的散斑图像盲去模糊方法。该方法以灰度直方图峰值的L 0范数与参考图像梯度域分布建立优化函数正则项,使用二次分裂方法估计清晰图像,再以交替迭代的方式进行卷积核细化。在模糊核估计完成后,使用Richardson-Lucy非盲去卷积方法完成散斑图像的复原。实验结果证明:所提出的散斑图像盲去模糊方法与针对自然图像与文本图像的经典方法相比,获得了更优的图像去模糊效果,并提高了数字图像相关测量精度与鲁棒性。     

基于APEX方法的湍流退化图像复原算法

在目标探测过程中,为了消除大气湍流带来的影响,提出了一种基于APEX方法的盲去卷积图像复原算法.该算法是一种非迭代的盲图像复原算法,以湍流退化系统具有G类点扩展函数为假设前提,通过模糊图像的频谱信息直接估计点扩展函数,并采用SECB方法实现目标图像的重建.本文对该算法的原理及其对湍流退化图像复原的可行性进行了深入研究,进行了真实的湍流退化图像的复原实验,其结果表明,该算法能够快速实现对湍流退化图像的重建,并具有一定的稳定性,能满足目标探测过程中的实时性要求.     

二阶广义总变分约束的太阳图像多帧盲解卷积

盲解卷积是常用的自适应光学图像事后重建方法之一。为提高盲解卷积对太阳(自适应光学)图像的重建效果，本文提出了基于二阶广义总变分的空变多帧盲解卷积算法。该算法首先利用交替最小化和半二次分裂方法求解本文提出的二阶广义总变分约束的空不变多帧盲解卷积模型；然后针对非等晕大视场太阳图像特性，利用重叠分块与加权拼接实现空变盲解卷积扩展。在一米新真空太阳望远镜(NVST)观测的真实太阳图像上进行的重建实验与分析表明，本文算法在主观视觉效果和客观指标上均具有较好的图像重建效果。     

卷积混合语音的联合块对角化盲源分离算法

针对语音卷积混合模型,提出了一种新的时域盲源分离算法。首先对观测信号进行重新排列,将卷积混合盲分离问题转化为瞬时混合盲分离问题,然后对联合近似对角化算法进行了推广,利用语音的非平稳和短时平稳特征定义联合差分相关矩阵和联合块对角化代价函数,通过鲁棒的白化过程和求解最优化问题实现卷积语音的盲分离。由于避免了时域卷积运算和变换域处理,使算法更加简单,复杂度更低。仿真结果验证了该算法的有效性,同时,就数据长度参数变化对信干比的影响,以及通过与基于线性预测的卷积盲分离算法和自然梯度卷积盲分离算法的比较对该算法的性能做了进一步的分析。     

基于APEX算法改进的图像复原算法

在高斯类点扩函数退化图像复原的研究中,提出了一种基于降晰图像频谱特征改进的APEX图像复原算法.该算法采用APEX算法的基本原理,根据图像频谱信息特征,对点扩散函数(PSF)估计过程进行了改进,采用加权最小二乘算法拟合出降晰图像频谱主方向,采用图像频谱主方向上的数据进行PSF估计,以利用更多的有效数据,从而减少PSF的估计误差.针对模拟和实际采集的降晰图像进行实验,采用主观视觉和峰值信噪比进行评价.实验结果表明,改进的算法较使用非主频谱方向上的频谱数据的复原算法在复原效果上有一定的提高.     

基于差分自相关的运动模糊图像尺度参数识别

在运动模糊图像盲复原的过程中,对点扩散函数(PSF)进行准确识别具有非常重要的意义.当识别出运动模糊方向后,二维PSF的识别问题就可以转化成一维PSF的识别,进而提出用一阶差分自相关的方法识别运动模糊尺度.给出了差分自相关方法用于尺度识别的理论分析,重点研究和推证了在该方法下噪声对模糊尺度识别的影响,提出了改进的差分自...     

基于Allen-Cahn方程图像修复的算子分裂方法（英文）

本文提出了一种基于Allen-Cahn方程图像修复的算子分裂方法.其核心思想是利用算子分裂方法将原问题分解为一个线性方程和一个非线性方程,线性方程使用有限差分CrankNicolson格式进行离散,非线性方程利用解析方法进行求解,因此时间和空间都能达到二阶精度.由于该方法只作用于图像需要修复的区域,而其余区域的像素值与原始图像的保持一样,可以大大提高计算效率.合成图像和真实图像的数值实验验证了该算法的正确性和有效性.     

改进的局部最小像素先验遥感图像盲复原算法

为了解决遥感图像盲复原时模糊核估计不准确、复原图像存在振铃效应的问题,提出改进的局部最小像素先验遥感图像盲复原算法。该算法首先引入极端通道先验与局部最小像素先验结合,对图像的强度进行更好的约束,有利于得到更好的潜在清晰图像;然后采用基于梯度的方法估计模糊核,模糊核估计与中间潜在清晰图像估计交替迭代进行,获得较为理想的模糊核;最后引入联合双边滤波器,采用改进的拉普拉斯与正则化图像复原算法抑制图像复原的振铃效应。实验结果表明,本文方法对遥感图像复原效果较好,恢复的图像边缘清晰,振铃伪影得到抑制且模糊核较为理想;客观评价指标峰值信噪比(PSNR)较前沿复原算法平均提高约1.40 dB,结构相似度(SSIM)平均提高约0.02。     

基于L_(1/2)/L_2正则化的图像盲去模糊方法

图像去模糊是图像识别和视频分析的基础性工作.在实际应用中,多数情形是模糊核未知的盲去模糊问题.盲去模糊问题是一个病态问题,通常需建立某种正则化模型求解.已有的图像去模糊正则化模型难以恢复模糊图像的细节,本文提出了一种基于L1/2/L2的正则化模型,并设计了求解该模型的交替投影迭代算法.数值实验表明,所提出的模型和算法能够更好地恢复模糊图像的细微结构,并且计算效率高,对参数的鲁棒性强.     

基于超拉普拉斯先验与核谱特性噪声图像盲去模糊

现有大部分盲图像去模糊方法对噪声敏感,即使少量的噪声可大大降低恢复图像的质量．考虑到模糊图像中同时隐含有清晰图像信息和模糊核信息,我们同时利用卷积核谱特性先验和清晰图像梯度域超拉普拉斯先验联合建立含噪图像盲去模糊模型,较单独使用卷积核先验与清晰图像先验建模更合理,也能获得更精确的估计图像．本文借助于Hessian矩阵,利用模糊图像及卷积核联合生成先验子,而非单独的估计图像先验子,建立优化模型．求解模型时,通过迭代策略交替细化模糊核和清晰图像．在清晰图像恢复阶段,因存在超拉普拉斯先验项,提出用变量分离法计算清晰图像．清晰图像采用快速傅里叶变换及封闭阈值公式求解,以提高优化速度．实验结果表明：与其他方法相比,本文方法能获得更鲁棒的模糊核和更精确的清晰图像,且收敛速度更快．     

Multiframe blind deconvolution of heavily blurred astronomical images

A multichannel blind deconvolution algorithm that incorporates the maximum-likelihood image restoration by several estimates of the differently blurred point-spread function (PSF) into the Ayers-Dainty iterative algorithm is proposed. The algorithm uses no restrictions on the image and the PSFs except for the assumption that they are positive. The algorithm employs no cost functions, input parameters, a priori probability distributions, or the analytically specified transfer functions. The iterative algorithm permits its application in the presence of different kinds of distortion. The work presents results of digital modeling and the results of processing real telescope data from several satellites. The proof of convergence of the algorithm to the positive estimates of object and the PSFs is given. The convergence of the Ayers-Dainty algorithm with a single processed frame is not obvious in the general case; therefore it is useful to have confidence in its convergence in a multiframe case. The dependence of convergence on the number of processed frames is discussed. Formulas for evaluating the quality of the algorithm performance on each iteration and the rule of stopping its work in accordance with this quality are proposed. A method of building the monotonically converging subsequence of the image estimates of all the images obtained in the iterative process is also proposed.     

Iterative statistical approach to blind image deconvolution

Image deblurring has long been modeled as a deconvolution problem. In the literature, the point-spread function (PSF) is often assumed to be known exactly. However, in practical situations such as image acquisition in cameras, we may have incomplete knowledge of the PSF. This deblurring problem is referred to as blind deconvolution. We employ a statistical point of view of the data and use a modified maximum a posteriori approach to identify the most probable object and blur given the observed image. To facilitate computation we use an iterative method, which is an extension of the traditional expectation-maximization method, instead of direct optimization. We derive separate formulas for the updates of the estimates in each iteration to enhance the deconvolution results, which are based on the specific nature of our a priori knowledge available about the object and the blur.     

Blind deconvolution of fluorescence micrographs by maximum-likelihood estimation

We report some recent algorithmic refinements and the resulting simulated and real image reconstructions of fluorescence micrographs by using a blind-deconvolution algorithm based on maximum likelihood estimation. Blind-deconvolution methods encompass those that do not require either calibrated or theoretical predetermination of the point-spread function (PSF). Instead, a blind deconvolution reconstructs the PSF concurrently with deblurring of the image data. Two-dimensional computer simulations give some definitive evidence of the integrity of the reconstructions of both the fluorescence concentration and the PSF. A reconstructed image and a reconstructed PSF from a two-dimensional fluorescent data set show that the blind version of the algorithm produces images that are comparable with those previously produced by a precursory nonblind version of the algorithm. They furthermore show a remarkable similarity, albeit not perfectly identical, with a PSF measurement taken for the same data set, provided by Agard and colleagues. A reconstructed image of a three-dimensional confocal data set shows a substantial axial smear removal. There is currently an existing trade-off in using the blind deconvolution in that it converges at a slightly slower rate than the nonblind approach. Future research, of course, will address this present limitation.     

Parametric blind deconvolution: a robust method for the simultaneous estimation of image and blur.

Blind-deconvolution microscopy, the simultaneous estimation of the specimen function and the point-spread function (PSF) of the microscope, is an underdetermined problem with nonunique solutions that are usually avoided by enforcing constraints on the specimen function and the PSF. We derived a maximum-likelihood-based method for blind deconvolution in which we assume a mathematical model for the PSF that depends on a small number of parameters (e.g., less than 20). The algorithm then estimates the unknown parameters together with the specimen function. The mathematical model ensures that all the constraints of the PSF are satisfied, and the maximum-likelihood approach ensures that the specimen is nonnegative. The method successfully estimates the PSF and removes out-of-focus blur. The PSF estimation is robust to aberrations in the PSF and to noise in the image.     

Accurate retrieval of region of interest for estimating point spread function and image deblurring

Extraction of an optimal region of interest (ROI) is crucial in many image processing applications, such as estimation of the point spread function (PSF) and blind deconvolution (BD). Although the amount of publications on PSF and BD is quite extensive; however, the work on ROI estimation has not received much attention. Existing methods which used heuristic models are not only time-consuming but also computationally expensive. In this paper, we proposed a new ROI retrieval scheme based on image partitioning and entropy measurement feedback. This method has low computation cost since it contains no matrix operations. Comprehensive experiments on real and synthetic datasets revealed that the proposed method is competitive when compared with existing search techniques, averaging at 26.1?dB, 0.46 and 1.44 on peak signal-to-noise ratio, universal image quality index and error ratio scales, respectively. On average, the proposed method takes less than 10?s to retrieve the ROI which is significantly faster compared to established solution.     

Upscaling image resolution of compact imaging systems using wavefront coding and a property of the point-spread function

We propose an image-resolution upscaling method for compact imaging systems. The image resolution is calculated using the resolving power of the optics and the pixel size of a digital image sensor. The resolution limit of the compact imaging system comes from its size and the number of allowed lenses. To upscale the image resolution but maintain the small size, we apply wavefront coding and image restoration. Conventional image restoration could not enhance the image resolution of the sensor. Here, we use the upscaled image of a wavefront-coded optical system and apply an image-restoration algorithm using a more precisely calculated point-spread function (PSF) as the deconvolution filter. An example of a wavefront-coded optical system with a 5-megapixel image sensor is given. The final image had a resolution equivalent to that of a 10-megapixel image using only four plastic lenses. Moreover, image degradation caused by hand motion could also be reduced using the proposed method.     

Parallel-distributed blind deconvolution based on a self-organizing neural network

A parallel-distributed blind deconvolution method based on a self-organizing neural network is introduced. A large degraded image is segmented into smaller subpatterns. Each subpattern can be used to get a blur function. Moreover, we propose a two-step unsupervised learning method in the self-organizing neural network. The two-step learning method includes parallel learning and series learning operations. The series learning operation is similar to a typical learning operation in the self-organizing neural network. The parallel learning operation is used as a positive perturbation to let the learning operation leave a local minimum. Several improved blur functions can be estimated from the different subpatterns, and the optimized blur function is evolved by use of a genetic algorithm. As the blur function is estimated, the source image of the large degraded image can be easily restored by use of a Wiener-type filter or other deconvolution methods. Computer simulations show that the proposed parallel-distributed blind deconvolution method gives good reconstruction and that the two-step learning method in the self-organizing neural network can promote learning. Since the main computational cost is dependent on the size of the subpattern, the proposed method is effective for the restoration of the large image.