引导滤波和对数变换算法融合的多尺度Retinex红外图像增强

针对采用红外成像仪获取红外图像边缘模糊、对比度差等缺点造成图像视觉效果差、质量低等问题.以多尺度Retinex算法为框架,依据引导滤波保边和梯度保持性,提出引导滤波和对数变换算法融合的多尺度Retinex红外图像增强方法.首先,用引导滤波替换MSR算法中的高斯滤波来估计照度分量.其次,将照度分量经过对数变换处理,执行低...     

改进直方图均衡和Retinex算法在灰度图像增强中的应用

为了提高光照条件变化下的图像增强效果,提出一种改进直方图均衡和Retinex算法的图像增强方法。对于待增强灰度图像,通过理想低通滤波获得图像低频分量,采用改进直方图均衡进行动态范围优化,利用引导图像滤波代替Retinex算法的高斯滤波对图像的高频分量进行估计,并对估计的结果进行线性放大处理。实验结果表明,相对单尺度和多尺度Retinex算法以及改进的直方图均衡化算法,本文方法从主观和客观评价方面都获得了更好的图像增强效果,有效提高了图像的视觉效果和可懂度,并具有较强的鲁棒性。     

尺度变化的Retinex红外图像增强

经典Retinex模型增强算法采用固定尺度高斯核平滑滤波,导致单一尺度Retinex无法进行全局有效增强,而多尺度Retinex权重系数选取困难,二者均不能满足视觉要求。针对以上问题,基于人眼视觉掩盖效应提出一种尺度变化高斯核平滑滤波的Retinex算法。首先利用人眼视觉掩盖效应的屏蔽函数检测像素邻域空间细节,依据像素区域细节信息丰富程度设计出尺度变化的高斯平滑滤波器,实现照度估计,最后对尺度变化高斯平滑滤波器实现提出实用方法。实验证明本文算法有效提高红外图像对比度,增强细节信息,在主观视觉效果和客观评价指标上整体优于修正对比度限制直方图均衡算法、单尺度Retinex、多尺度Retinex及平稳小波和Retinex增强算法。     

基于双域分解的夜间车辆红外图像研究

夜间车辆交通红外图像光照不均,导致车辆图像细节纹理较弱,识别难度较大。为此,提出基于双域分解的夜间车辆交通红外偏振图像增强方法。采用改进Retinex低照度图像光照补偿算法,分解图像为低频图像与高频图像,对低频图像去雾、优化其对比度,对高频图像去噪与增强,合成低频、高频图像实现夜间车辆交通红外偏振图像增强。实验测试结果证明,对比传统方法,所提方法增强后图像亮度与对比度得以优化,且细节信息更丰富,具有理想的视觉效果。     

基于双残差卷积网络的低照度图像增强

为解决当前低照度图像增强问题,提出了一种基于双残差卷积网络的图像增强算法.首先,根据Retinex理论模型,将正常照度图像合成低照度图像,再分别将它们分解在R(红)、G(绿)、B(蓝)3个分量上,然后通过特征提取模块和双残差模块学习低照度图像与正常照度图像在各分量的映射关系,获得各分量上的增强图像,最后合成增强的RGB...     

采用深度学习与图像融合混合实现策略的低照度图像增强算法

提出了一种采用深度学习与图像融合混合实现策略的低照度图像增强算法.首先,利用照射分量预测模型直接基于输入的低照度图像快速地估计出其最佳照射分量并在Retinex模型框架下获得一张整体上适度曝光图像;其次,将低照度图像本身及它的过曝光图像作为适度曝光图像的修正补充图像参与融合;最后,采用局部结构化融合和色度加权融合机制技...     

基于Retinex的一种图像去雾算法

由于中心环绕Retinex图像增强算法尺度的选择有限,不能在对有雾天气下采集的图像进行有效去雾的同时增强其细节,提出一种基于Retinex算法的曲波变换图像增强算法.首先,根据Retinex算法,先用高斯函数估计出图像的入射分量,再通过Retinex算法将图像的反射分量得出,然后利用曲波变换的多尺度特性将反射分量进行子带分解,对高频子带采用自动变换阈值处理,对低频子带采用线性拉伸,增强其对比度,最后将曲波系数进行合成,得出处理后的图像.实验结果表明,用该方法对雾天图像处理后,图像的纹理细节更清晰,信噪比和信息熵明显提高,视觉效果改善,失真度显著减小.     

基于卷积分析稀疏表示和相位一致性的低照度图像增强

针对低照度图像存在的对比度低、视觉效果差等问题,提出一种基于卷积分析稀疏表示和相位一致性的低照度图像增强方法.该方法基于Retinex模型,在估计照度图像时采用卷积分析稀疏表示进行约束,所用滤波器一部分人工设定,一部分由样本训练自动获得;在计算反射图像时利用单演相位一致性特征,施加相位一致性残余最小约束来恢复细节;通过联合约束并进行优化,得到的反射图像即为最终的增强结果.对大量低照度图像进行实验,并与当前先进方法相比,结果表明,本文方法不仅提高了图像的亮度与对比度,增强了细节,而且在多个客观评价指标上都优于其他方法.     

基于双重迭代的零样本低照度图像增强

针对低光照条件下拍摄图像质量低下的问题,该文提出一种基于双重迭代的零样本低照度图像增强方法。其外层迭代通过卷积神经网络估计增强参数,再由内层迭代进行图像增强,增强结果进一步用于计算损失函数并反馈更新外层的参数估计网络,最终通过多轮迭代生成高质量的图像。在该框架下,还设计了多尺度增强系数估计模块、基于注意力的像素级大气光估计模块,并提出了基于亮度对比度、大气光、颜色均衡以及图像平滑性先验的无监督损失函数。大量实验结果表明,该方法可有效将低光照图像增强为高质量的清晰图像,其性能优于现有的同类方法。同时该方法基于零样本学习,不需任何训练数据集,具有良好的普适性。     

介质平面光波导TE0模模场分布的高斯近似

基于介质平面光波导端面无受限衍射场光束的光束传输因子的特点,阐明光波导TE0模模场分布采用高斯分布近似表达的合理性。基于场分布间的匹配效率计算公式,提出采用等效匹配效率方法确定用于高斯近似表达的等效模场半宽度,给出基于光波导芯层半宽度和归一化芯层驻波参量和归一化包层倏逝波参量表达的光波导等效模场半宽度的函数表达式,给出高斯近似分布与光波导本征场分布的匹配效率,阐明采用等效匹配效率方法确定等效模场半宽度的合理性。采用求解方程组的方法,给出基于光波导芯层半宽度和归一化频率表达的光波导等效模场半宽度的拟合函数表达式,并基于拟合引起的误差分析阐明了拟合函数表达式的精确性。     

Tomographic fluorescence imaging in tissue phantoms: a novel reconstruction algorithm and imaging geometry

A novel image reconstruction algorithm has been developed and demonstrated for fluorescence-enhanced frequency-domain photon migration (FDPM) tomography from measurements of area illumination with modulated excitation light and area collection of emitted fluorescence light using a gain modulated image-intensified charge-coupled device (ICCD) camera. The image reconstruction problem was formulated as a nonlinear least-squares-type simple bounds constrained optimization problem based upon the penalty/modified barrier function (PMBF) method and the coupled diffusion equations. The simple bounds constraints are included in the objective function of the PMBF method and the gradient-based truncated Newton method with trust region is used to minimize the function for the large-scale problem (39919 unknowns, 2973 measurements). Three-dimensional (3-D) images of fluorescence absorption coefficients were reconstructed using the algorithm from experimental reflectance measurements under conditions of perfect and imperfect distribution of fluorophore within a single target. To our knowledge, this is the first time that targets have been reconstructed in three-dimensions from reflectance measurements with a clinically relevant phantom.     

引导滤波与像素重分布的低照图像增强

针对现有的图像增强方法存在欠增强、过增强以及对比度低等缺陷,提出了一种引导滤波与像素重分布的低照图像增强算法。方法充分利用引导滤波的边缘保持特性,用引导滤波对光照图像进行估计,然后对光照图像的像素进行相对均匀重分布,全面提升光照图像的亮度与对比度。最后将像素重分布增强处理后的光照图像和反射图像作反Retinex变换,得到最后的增强图像。实验结果证明,相对现有的图像增强方法,本文方法具有更优的图像增强效果,图像对比度与纹理结构更清晰。     

DSE-Net: Deep simultaneous estimation network for low-light image enhancement

This paper presents a novel approach for low-light image enhancement. We propose a deep simultaneous estimation network (DSE-Net), which simultaneously estimates the reflectance and illumination for low-light image enhancement. The proposed network contains three modules: image decomposition, illumination adjustment, and image refinement module. The DSE-Net uses a novel branched encoder–decoder based image decomposition module for simultaneous estimation. The proposed decomposition module uses a separate decoder to estimate illumination and reflectance. DSE-Net improves the estimated illumination using the illumination adjustment module and feeds it to the proposed refinement module. The image refinement module aims to produce sharp and natural-looking output. Extensive experiments conducted on a range of low-light images demonstrate the efficacy of the proposed model and show its supremacy over various state-of-the-art alternatives.     

A low light natural image statistical model for joint contrast enhancement and denoising

We study the problem of joint low light image contrast enhancement and denoising using a statistical approach. The low light natural image in the band pass domain is modeled by statistically relating a Gaussian scale mixture model for the pristine image, to the low light image, through a detail loss coefficient and Gaussian noise. The detail loss coefficient is statistically described using a posterior distribution with respect to its estimate based on a prior contrast enhancement algorithm. We then design our low light enhancement and denoising (LLEAD) method by computing the minimum mean squared error estimate of the pristine image band pass coefficients. We create the Indian Institute of Science low light image dataset of well-lit and low light image pairs to learn the model parameters and evaluate our enhancement method. We show through extensive experiments on multiple datasets that our method helps better enhance the contrast while simultaneously controlling the noise when compared to other state of the art joint contrast enhancement and denoising methods.     

视觉检测系统LED光源照度优化方法

针对当前视觉检测系统LED光源照度优化研究中存在的照度效果评价因素单一、照度优化方法通用性不足等问题,以芯片封装质量视觉检测为例,提出一种基于改进樽海鞘算法的LED光源照度优化方法。该方法在单个LED光源照度数学模型基础上,建立标准条形LED阵列光源照度数学模型,获取条形LED阵列在任意空间位姿与被测面的照度值;基于照度均匀度、照度梯度变化与对中度、平均照度、目标与背景区分度等因素建立平面照度效果评价函数;提出改进樽海鞘算法,通过改进算法收敛系数、速度、领导者与追随者位置等更新策略,增强区域搜索的多样性;应用改进樽海鞘算法对平面照度效果评价函数进行优化求解,获取具有最优照度效果的空间位姿参数。实验结果表明:考察优化区域的相对照度分布,文中提出的LED光源照度优化方法所得照度分布与实际测量所得照度分布结果基本一致,目标区域理论照度均匀度在98.78%以上,误差在5.57%以内。因此文中提出方法优化目标合理,可用于视觉检测系统具有最优照度效果时光源位姿信息参数的获取。     

基于三角函数变换与IRDPSO优化的图像增强算法

针对复杂环境下如阴天、雾天、夜晚、光照较弱等条件下拍摄的图像存在对比度不足、整体偏暗等问题,提出了一种基于三角函数变换与改进随机漂移粒子群算法的图像增强算法。该图像增强方法主要包括四个步骤,首先将彩色图像转换为灰度图像,然后对灰度图像利用三角函数变换提高对比度,然后再对图像进行拉布拉斯算子增强,最后再对图像进行色彩恢复。为了提高算法的自适应性,针对三角函数变换中的参数、以及拉布拉斯算子模板的参数选择问题,将改进随机漂移粒子群算法(IRDPSO)与图像增强算法结合,利用信息熵和图像标准差构造适应度函数,对参数进行寻优。将该方法与其他四种算法进行比较。实验结果表明:文中算法简单,增强后的图像信息熵值、标准差值均较大,图像颜色失真度小,增强效果均比其他几种算法好,提高了图像的质量和对比度。     

Constrained nonlinear optimization approaches to color-signalseparation

Separating a color signal into illumination and surface reflectance components is a fundamental issue in color reproduction and constancy. This can be carried out by minimizing the error in the least squares (LS) fit of the product of the illumination and the surface spectral reflectance to the actual color signal. When taking in account the physical realizability constraints on the surface reflectance and illumination, the feasible solutions to the nonlinear LS problem should satisfy a number of linear inequalities. Four distinct novel optimization algorithms are presented to employ these constraints to minimize the nonlinear LS fitting error. The first approach, which is based on Ritter's superlinear convergent method (Luengerger, 1980), provides a computationally superior algorithm to find the minimum solution to the nonlinear LS error problem subject to linear inequality constraints. Unfortunately, this gradient-like algorithm may sometimes be trapped at a local minimum or become unstable when the parameters involved in the algorithm are not tuned properly. The remaining three methods are based on the stable and promising global minimizer called simulated annealing. The annealing algorithm can always find the global minimum solution with probability one, but its convergence is slow. To tackle this, a cost-effective variable-separable formulation based on the concept of Golub and Pereyra (1973) is adopted to reduce the nonlinear LS problem to be a small-scale nonlinear LS problem. The computational efficiency can be further improved when the original Boltzman generating distribution of the classical annealing is replaced by the Cauchy distribution.     

Self-organized night video enhancement for surveillance systems

Night video enhancement techniques are widely used for identifying suspicious activities captured by night visual surveillance systems. However, artificial light sources present in the surroundings deteriorate the visual quality of the video captured during night. This non-uniform illumination reduces the object identification and tracking capability of a real-time visual security system. Thus, a uniform enhancement technique is insufficient for handling such uneven illumination. In this paper, we propose a novel night video enhancement scheme based on a hierarchical self-organizing network. This proposed scheme automatically groups and enhances the neighboring pixels of dark and light regions in each frame. In this scheme, two-level self- organizing neural networks were hierarchically arranged to group similar pixels present in the night video frame. We applied the no-reference-based performance evaluation metrics for measuring the objective quality of the video. The experimental results showed that our proposed approach considerably enhances the visual perception of the video captured at night under varied illumination conditions.     

Hyperparameter Optimization for Machine Learning Models Based on Bayesian Optimization

Hyperparameters are important for machine learning algorithms since they directly control the behaviors of training algorithms and have a significant effect on the performance of machine learning models. Several techniques have been developed and successfully applied for certain application domains. However, this work demands professional knowledge and expert experience. And sometimes it has to resort to the brute-force search. Therefore, if an efficient hyperparameter optimization algorithm can be developed to optimize any given machine learning method, it will greatly improve the efficiency of machine learning. In this paper, we consider building the relationship between the performance of the machine learning models and their hyperparameters by Gaussian processes. In this way, the hyperparameter tuning problem can be abstracted as an optimization problem and Bayesian optimization is used to solve the problem. Bayesian optimization is based on the Bayesian theorem. It sets a prior over the optimization function and gathers the information from the previous sample to update the posterior of the optimization function. A utility function selects the next sample point to maximize the optimization function. Several experiments were conducted on standard test datasets. Experiment results show that the proposed method can find the best hyperparameters for the widely used machine learning models, such as the random forest algorithm and the neural networks, even multi-grained cascade forest under the consideration of time cost.