基于单目和双目视觉信息的全参考立体图像质量评价模型

在立体图像质量评价领域,有效地模拟人类视觉系统对图像质量进行评价具有重要意义,考虑到人眼的视觉感知特性,基于单目和双目视觉信息构建一种立体图像质量评价模型MB-FR-SIQA。采用基于结构相似性的立体视差算法得到参考和失真立体图像的视差矩阵,结合Gabor能量响应图、显著性图和视差矩阵生成中间视图,并优化左右眼加权系数计算方法,以提高生成中间视图的准确性。分别利用单目图像和中间视图提取单目和双目视觉信息,计算单目质量分数和双目质量分数,并融合得到立体图像的质量分数,达到评价立体图像质量的目的。实验结果表明,MB-FR-SIQA模型在LIVE-I数据库上具有较高的预测精度,其斯皮尔曼等级相关系数、皮尔森线性相关系数、均方根误差分别为0.945、0.951、5.318,且预测的质量分数符合人类主观评估。     

基于空域自然场景统计的无参考立体图像质量评价模型

针对现有的评价方法大都将图像变换到不同的坐标域问题,提出一种基于空域自然场景统计(NSS)的通用型无参考立体图像质量评价模型。在评价中为了更好地结合人类双目视觉特性, 将左右图像融合成一幅独眼图;评价模型首先统计独眼图归一化亮度(CMSCN)系数分布规律,进而对独眼图提取空域自然场景统计特征;其次,统计视差图归一化亮度(DMSCN)系数的分布规律,并对用光流法得到的视差图提取同样的特征;最后,通过支持向量回归(SVR)建立立体图像特征信息与主观评价值(DMOS)之间的关系,从而预测得到图像质量的客观评价值。实验结果表明,该评价模型对立体数据测试库进行评价,其Pearson线性相关系数(PLCC)和Spearman等级相关系数(SROCC)值均在0.94以上;对于非对称立体图像库,PLCC和SROCC值分别接近0.91和0.93。该模型能够很好地预测人眼对立体图像的主观感知。     

Quality assessment of perceptual color video based on a top-down framework and quaternion

Objective video quality assessment is of great importance in a variety of video processing applications. Most existing video quality metrics either focus primarily on capturing spatial artifacts in the video signal, or are designed to assess only grayscale video thereby ignoring important chrominance information. In this paper, on the basis of the top-down visual analysis of cognitive understanding and video features, we propose and develop a novel full-reference perceptual video assessment technique that accepts visual information inputs in the form of a quaternion consisting of contour, color and temporal information. Because of the more important role of chrominance information in the “border-to-surface” mechanism at early stages of cognitive visual processing, our new metric takes into account the chrominance information rather than the luminance information utilized in conventional video quality assessment. Our perceptual quaternion model employs singular value decomposition (SVD) and utilizes the human visual psychological features for SVD block weighting to better reflect perceptual focus and interest. Our major contributions include: a new perceptual quaternion that takes chrominance as one spatial feature, and temporal information to model motion or changes across adjacent frames; a three-level video quality measure to reflect visual psychology; and the two weighting methods based on entropy and frame correlation. Our experimental validation on the video quality experts’ group (VQEG) Phase I FR-TV test dataset demonstrated that our new assessment metric outperforms PSNR, SSIM, PVQM (P8) and has high correlation with perceived video quality.     

基于眼优势的非对称失真立体图像质量评价

针对现有立体图像质量评价算法对非对称失真立体图像的评价准确性及执行效率较低的问题,提出一种基于眼优势的非对称失真立体图像质量评价算法.首先采用梯度幅值响应来模拟左右眼输入的刺激强度,并根据人类视觉系统的眼优势原理分别以左和右视点图像作为主视图合成两幅融合图像;其次,利用旋转不变统一局部二值模式直方图、皮尔逊线性相关系数以及非对称广义高斯模型,获取左右融合图像以及左右梯度幅值响应图像中的多种能够反映立体图像质量好坏的特征;最后,利用自适应增强的支持向量回归模型将感知特征向量映射为图像质量值.在四个基准测试数据库上的实验结果表明:本文所提出算法大幅提升了非对称失真立体图像的评价准确性,且具有较高的执行效率.这些优势说明本文算法所提取的特征描述能力更强,质量映射模型的稳定性更好.     

多核学习纹理特征的立体图像质量评价

为了有效地评价各种失真类型双目立体图像的质量,提出利用多核学习机学习立体图像平面纹理信息和3D映射信息的通用无参考立体图像质量评价IQA方法。该方法首先利用立体匹配模型对左右视图进行处理,获得相应的视差图DM和误差能量图DMEE;对左右视图、视差图和误差能量图进行相位一致性和结构张量变换,获得它们的平坦区和边缘区;分别提取左右视图两个区域纹理特征作为平面信息,提取视差图的纹理特征和误差能量图的统计特征作为3D信息;将所有特征作为多核学习机的输入,利用多核学习的信息融合能力预测待测失真立体图像质量。由于充分利用了立体图像的左右视图、视差图和误差能量图的失真信息,以及多核学习的信息融合能力,该方法具有很好的前景。在LIVE 3D图像质量数据库上的实验表明,该方法与主观质量有较高一致性,与现有的双目立体质量评价方法相比有很大的竞争力。     

基于四元数的改进型最小核值相似区边缘检测

针对基于亮度和色度的彩色图像边缘检测在检测过程中忽略亮度和色度之间关联性而导致部分边缘不能有效地被检测出来的问题,提出了一种基于四元数的改进型最小核值相似区(SUSAN)边缘检测算法。首先,利用四元数矢量旋转原理将HSI颜色空间的三维信息映射成二维平面信息实现空间降维,同时引入标量V来综合表示H、S、I三通道之间的关系;然后,将标量V作为算子的核函数;最后,利用改进的SUSAN算子完成图像的边缘检测。实验结果表明,提出的算法针对色度相同、饱和度存在差异以及饱和度相同、色度存在差异的彩色图像,在边缘检测的定位误差率上降低了1.5%。在实际的应用中,能够更好地获得图像中的目标信息,同时也为后续的分割和识别研究提供更好的先验知识。     

基于小波变换的水下降质图像复原算法

根据水下成像的物理模型,提出一种基于小波变换的水下降质图像清晰化处理算法。该算法将RGB图像转换为YUV图像,根据图像的对比度,对亮度Y图像利用小波变换自适应估计介质散射光的大小,增强水下降质图像的对比度,并在小波变换的低频子带上进行非线性亮度调节,消除水下图像的光照不均问题,将亮度Y图像的处理结果与颜色分量U、V合成得到清晰的水下彩色图像。实验结果表明,该算法可以自适应实现水下观测图像的清晰化处理。     

Quaternion wavelet phase based stereo matching for uncalibrated images

This paper proposes a quaternion wavelet phase based stereo matching (QWPSM) scheme for uncalibrated image pairs. In this scheme, we estimate the disparity by directly establishing correspondences between quaternionic phase structures of two quaternion wavelet filtered (QWF) images. Firstly, linear-phase quaternion wavelet filters (LPQWFs) are constructed from real biorthogonal wavelet bases. Then, quaternion phases are extracted under each scale through quaternion wavelet filtering of the multiscale transformed image pyramids. The disparity estimation is formed as a minimization process of a local energy weighted cost function, and propagated from coarse to fine scales. Costs can adaptively alleviate the negative effects of phase singularities, which are the main causes of mismatches in phase-based stereo matching. Multiscale matching strategy is used to avoid phase wrapping and improve convergence speed. Experimental results are promising in various image pairs.     

考虑双目竞争视觉现象的非对称失真立体图像质量评价方法

目的 现有方法存在特征提取时间过长、非对称失真图像预测准确性不高的问题,同时少有工作对非对称失真与对称失真立体图像的分类进行研究,为此提出了基于双目竞争的非对称失真立体图像质量评价方法。方法 依据双目竞争的视觉现象,利用非对称失真立体图像两个视点的图像质量衰减程度的不同,生成单目图像特征的融合系数,融合从左右视点图像中提取的灰度空间特征与HSV （hue-saturation-value）彩色空间特征。同时,量化两个视点图像在结构、信息量和质量衰减程度等多方面的差异,获得双目差异特征。并且将双目融合特征与双目差异特征级联为一个描述能力更强的立体图像质量感知特征向量,训练基于支持向量回归的特征—质量映射模型。此外,还利用双目差异特征训练基于支持向量分类模型的对称失真与非对称失真立体图像分类模型。结果 本文提出的质量预测模型在4个数据库上的SROCC （Spearman rank order correlation coefficient）和PLCC （Pearson linear correlation coefficient）均达到0.95以上,在3个非对称失真数据库上的均方根误差（root of mean square error,RMSE）取值均优于对比算法。在LIVE-II（LIVE 3D image quality database phase II）、IVC-I（Waterloo-IVC 3D image qualityassessment database phase I）和IVC-II （Waterloo-IVC 3D image quality assessment database phase II）这3个非对称失真立体图像测试数据库上的失真类型分类测试中,对称失真立体图像的分类准确率分别为89.91%、94.76%和98.97%,非对称失真立体图像的分类准确率分别为95.46%,92.64%和96.22%。结论 本文方法依据双目竞争的视觉现象融合左右视点图像的质量感知特征用于立体图像质量预测,能够提升非对称失真立体图像的评价准确性和鲁棒性。所提取双目差异性特征还能够用于将对称失真与非对称失真立体图像进行有效分类,分类准确性高。     

基于视差图和复数轮廓波变换的无参考立体图像质量评价

现有的2D图像质量评价方法并不能很好地应用于立体图像质量评价中。为了有效评价不同失真立体图像的质量,提出了一种基于视差图和复数轮廓波变换的无参考图像质量评价方法。首先提取了能够反映3D信息的视差图,然后对左右失真图像和视差图进行复数轮廓波变换,计算能量和能量差特征,最后通过支持向量回归SVR模型训练学习,预测图像质量分数。实验结果表明,此方法优于当前文献报道的立体图像质量评价方法。     

Visual comfort in 2D/3D modes of glasses‐type stereoscopic displays

In this study, we compared visual comfort in 2D/3D modes of the pattern retarder (PR) and shutter glasses (SG) stereoscopic displays by changing viewing factors and image contents. The viewing factors include ambient illuminance/monitor luminance/background luminance and image contents mainly are determined with different disparity limits. The degrees of 2D/3D visual comfort were investigated by using various combinations of ambient illuminance, monitor luminance, background luminance, and disparity limit. A series of psychological experiments were also performed to compare 2D and 3D viewing experiences for the passive PR and active SG stereoscopic displays and to discover more comfortable conditions under various variable combinations. The experiment results show that the various variable combinations affecting visual comfort in the passive PR and active SG stereoscopic displays were significantly different. Finally, we suggest more comfortable conditions of viewing 2D and 3D images for the PR and SG stereoscopic displays.     

An HVS based adaptive quantization scheme for the compression of color images

In this paper a Human Visual System based adaptive quantization scheme is proposed. The proposed algorithm supports perceptually lossless as well as lossy compression. The algorithm uses a transform based compression approach using the wavelet transform, and has incorporated vision models for the compression of both luminance and chrominance components. The major strength of the coder is the incorporation of the vision model for the chrominance components and the optimum way in which the scales are distributed among the luminance and chrominance components to achieve higher compression ratios. The perceptual model developed for the color components gives flexibility for giving more compression for the color components without causing any color degradations. For each image the visual thresholds are evaluated and an optimum bit allocation is done in such a way that the quantization error is always less than the visual distortion for the given rate. To validate the strength of the proposed algorithm, the perceptual quality of the images reconstructed using the proposed coder is compared with the images reconstructed with JPEG2000 standard coder, for the same compression. To evaluate the perceptual quality of the compressed images latest perceptual quality matrices such as Structural Similarity Index, Visual Information Fidelity and Visual Signal-to-Noise Ratio are used. The results obtained reveal that the proposed structure gives excellent improvement in perceptual quality compared to the existing schemes, for both lossy as well as lossless compression. These advantages make the proposed algorithm a good candidate for replacing the quantizer stage of the current image compression standards.     

基于视差重映射的立体图像视觉舒适度提升

目的 针对人眼观看立体图像内容可能存在的视觉不舒适性,基于视差对立体图像视觉舒适度的影响,提出了一种结合全局线性和局部非线性视差重映射的立体图像视觉舒适度提升方法。方法 首先,考虑双目融合限制和视觉注意机制,分别结合空间频率和立体显著性因素提取立体图像的全局和局部视差统计特征,并利用支持向量回归构建客观的视觉舒适度预测模型作为控制视差重映射程度的约束;然后,通过构建的预测模型对输入的立体图像的视觉舒适性进行分析,就欠舒适的立体图像设计了一个两阶段的视差重映射策略,分别是视差范围的全局线性重映射和针对提取的潜在欠舒适区域内视差的局部非线性重映射;最后,根据重映射后的视差图绘制得到舒适度提升后的立体图像。结果 在IVY Lab立体图像舒适度测试库上的实验结果表明,相较于相关有代表性的视觉舒适度提升方法对于欠舒适立体图像的处理结果,所提出方法在保持整体场景立体感的同时,能更有效地提升立体图像的视觉舒适度。结论 所提出方法能够根据由不同的立体图像特征构建的视觉舒适度预测模型来自动实施全局线性和局部非线性视差重映射过程,达到既改善立体图像视觉舒适度、又尽量减少视差改变所导致的立体感削弱的目的,从而提升立体图像的整体3维体验。     

Local MAP estimation for quality improvement of compressed color images

This paper concerns color image restoration aiming at objective quality improvement of compressed color images in general rather than merely artifact reduction. In compressed color images, colors are usually represented by luminance and chrominance components. Considering characteristics of human vision system, chrominance components are generally represented more coarsely than luminance component. To recover such chrominance components, we previously proposed a model-based chrominance restoration algorithm where color images are modeled by a Markov random field. This paper presents a color image restoration algorithm derived by the MAP estimation, where all components are totally estimated. Experimental results show that the proposed restoration algorithm is more effective than the previous one.     

Perceived quality measurement of stereoscopic 3D images based on sparse representation and binocular combination

Measurement of the perceived quality of stereoscopic three-dimensional (S3D) images has attracted an increasing amount of research interest in recent years. This paper proposes a S3D image quality measurement (IQM) metric based on sparse representation and binocular combination. The proposed method involves learning binocular and monocular dictionaries from a training database such that the sparse features of binocular combination can be expressed by a linear combination of a few selected basis feature vectors. Following this, scores for the similarity of these sparse features between reference and distorted S3D images are measured. Based on the observation that sparse features are invariant against weak degradations, similarity scores of the features of the gradient magnitude of binocular combination are then computed and used as a complementary feature. Finally, by using kernel-based support vector regression (SVR), these similarity scores are integrated into an overall quality value. Experimental results on three public S3D-IQM datasets show that in comparison with the relevant existing metrics, the devised metric attains significantly high consistency alignment with subjective quality assessment.     

基于HVS和交叉掩蔽的彩色图像编码算法

以层树分集（SPIHT）编码方案为基础,结合人类视觉系统（HVS）模型和人类视觉对彩色图像分量亮度和色度的不同敏感性,提出了一种基于非对称编码和交叉掩蔽的小波域彩色图像压缩编码算法。该算法首先将原始图像从RGB空间转换到YCbCr空间,然后对YCbCr空间的各分量进行离散小波变换;之后根据人类视觉对彩色图像的亮度分量的敏感性,用交叉掩蔽模型对亮度分量的小波系数进行加权处理;与此同时,利用非对称编码和SPIHT编码思想完成图像的压缩。仿真实验结果表明,文中算法是一种高效的图像压缩编码方法,其压缩效果明显优于SPIHT编码方案。     

彩色立体图像质量评价方法

现有的大多数立体图像质量评价方法都是将彩色图像转换为灰度图像,从而丧失了色彩信息,不利于对彩色立体图像作出正确评价,针对这一问题,提出了一种彩色立体图像质量评价方法。首先,通过对参考图像对和失真图像对分别进行主成分分析(PCA)融合生成彩色图像,利用彩色小波变换分别提取彩色融合图像的低频系数;然后,把低频系数信息用四元数表示,即将低频系数的色相分量局部均值作为四元数的实部,三基色分量作为四元数的虚部,通过四元数奇异值分解得到奇异值特征向量;最后,对参考图像和失真图像的奇异值特征向量作余弦夹角、巴氏距离、卡方距离,分别作为立体图像质量评价指标。该方法在德克萨斯大学公布的对称失真立体图像库和非对称失真立体图像库分别进行验证,线性相关系数和斯皮尔曼等级相关系数(SROCC)在对称失真库中可高达0.919和0.923,与主观评价吻合度很高。     

Robust and transparent watermarking scheme for colour images

In this study, a robust and transparent watermarking scheme for colour images is proposed. The colour features for the human visual system are utilised to design the colour watermarking scheme. Through the exploitation of the perceptual redundancy of colour images, the proposed watermarking scheme is perceptually tuned to embed and detect the watermark in the perceptually significant sub-bands of luminance and chrominance components of colour images in the wavelet domain. The employment of the uniformity in the uniform colour space and the masking effect mainly due to local variations in luminance magnitude leads to that the perceptual redundancy of colour images can be measured. By using the estimated perceptual redundancy in the form of error visibility thresholds of wavelet coefficients of the colour image, high strength watermarks are invisibly embedded into coefficients of the host colour image for resisting compression and malicious attacks. Simulation results show that the estimation of perceptual redundancy is helpful to the design of the watermarking scheme for colour images. The performance in terms of robustness and transparency of the proposed watermarking scheme is superior to that of the existing scheme.     

深度学习单目深度估计研究进展

单目深度估计是从单幅图像中获取场景深度信息的重要技术,在智能汽车和机器人定位等领域应用广泛,具有重要的研究价值。随着深度学习技术的发展,涌现出许多基于深度学习的单目深度估计研究,单目深度估计性能也取得了很大进展。本文按照单目深度估计模型采用的训练数据的类型,从3个方面综述了近年来基于深度学习的单目深度估计方法:基于单图像训练的模型、基于多图像训练的模型和基于辅助信息优化训练的单目深度估计模型。同时,本文在综述了单目深度估计研究常用数据集和性能指标基础上,对经典的单目深度估计模型进行了性能比较分析。以单幅图像作为训练数据的模型具有网络结构简单的特点,但泛化性能较差。采用多图像训练的深度估计网络有更强的泛化性,但网络的参数量大、网络收敛速度慢、训练耗时长。引入辅助信息的深度估计网络的深度估计精度得到了进一步提升,但辅助信息的引入会造成网络结构复杂、收敛速度慢等问题。单目深度估计研究还存在许多的难题和挑战。利用多图像输入中包含的潜在信息和特定领域的约束信息,来提高单目深度估计的性能,逐渐成为了单目深度估计研究的趋势。     

No-reference image quality assessment using modified extreme learning machine classifier

In this paper, we present a machine learning approach to measure the visual quality of JPEG-coded images. The features for predicting the perceived image quality are extracted by considering key human visual sensitivity (HVS) factors such as edge amplitude, edge length, background activity and background luminance. Image quality assessment involves estimating the functional relationship between HVS features and subjective test scores. The quality of the compressed images are obtained without referring to their original images (‘No Reference’ metric). Here, the problem of quality estimation is transformed to a classification problem and solved using extreme learning machine (ELM) algorithm. In ELM, the input weights and the bias values are randomly chosen and the output weights are analytically calculated. The generalization performance of the ELM algorithm for classification problems with imbalance in the number of samples per quality class depends critically on the input weights and the bias values. Hence, we propose two schemes, namely the k-fold selection scheme (KS-ELM) and the real-coded genetic algorithm (RCGA-ELM) to select the input weights and the bias values such that the generalization performance of the classifier is a maximum. Results indicate that the proposed schemes significantly improve the performance of ELM classifier under imbalance condition for image quality assessment. The experimental results prove that the estimated visual quality of the proposed RCGA-ELM emulates the mean opinion score very well. The experimental results are compared with the existing JPEG no-reference image quality metric and full-reference structural similarity image quality metric.