Effects of luminance and colour contrast on the search of information on display devices

For black-and-white alphanumeric information, the speed of visual perception decreases with decreasing contrast. We investigated the effect of luminance contrast on the speed of visual search and reading when characters and background differed also with respect to colour. The luminance contrast between background and characters was varied, while colour contrast was held nearly constant. Stimuli with moderate (green/grey) or high colour contrast (green/red or yellow/blue), and three character sizes (0.17, 0.37, and 1.26 deg) were used. Eye movements were recorded during the visual search task. We found that the visual search times, number of eye fixations, and mean fixation durations increased strongly with decreasing luminance contrast despite the presence of colour contrast. The effects were largest for small characters (0.17 deg), but occurred also for medium (0.37 deg), and in some cases for large (1.26 deg) characters. Similarly, reading rates decreased with decreasing luminance contrast. Thus, moderate or even high colour contrast does not guarantee quick visual perception, if the luminance contrast between characters and background is small. This is probably due to the fact that visual acuity (the ability to see small details) is considerably lower for pure colour information than for luminance information. Therefore, in user interfaces, good visibility of alphanumeric information requires clear luminance (brightness) difference between foreground and background.     

多模态特征融合的视频记忆度预测

随着网络视频的爆炸式增长,视频记忆度成为热点研究方向。视频记忆度是衡量一个视频令人难忘的程度指标,设计自动预测视频记忆度的计算模型有广泛的应用和前景。当前对视频记忆度预测的研究多集中于普遍的视觉特征或语义因素,没有考虑深度特征对视频记忆度的影响。着重探索了视频的深度特征,在视频预处理后利用现有的深度估计模型提取深度图,将视频原始图像和深度图一起输入预训练的ResNet152网络来提取深度特征;使用TF-IDF算法提取视频的语义特征,并对视频记忆度有影响的单词赋予不同的权重;将深度特征、语义特征和从视频内容中提取的C3D时空特征进行后期融合,提出了一个融合多模态的视频记忆度预测模型。在MediaEval 2019会议提供的大型公开数据集（VideoMem）上进行实验,在视频的短期记忆度预测任务中达到了0.545（长期记忆度预测任务：0.240）的Spearman相关性,证明了该模型的有效性。     

动态场景下结合语义的半直接法视觉里程计

针对传统视觉里程计方法在动态环境中跟踪效果差、容易失效的问题,提出了一种适用于动态场景下融合语义的半直接法视觉里程计,主要由语义分割、位姿跟踪和地图管理三部分组成。首先,使用语义分割网络Mask R-CNN对每一帧图像进行语义分割,得到语义分割二值图,结合语义先验信息删除动态特征;然后,利用静态部分特征进行帧间对齐和特征匹配完成位姿跟踪;最后,通过筛选关键帧和更新3D点深度对地图更新,并利用光束平差算法对局部地图的位姿和3D点进行优化。在公共数据集TUM RGB-D的动态序列集上的实验结果表明,该方法与ORB-SLAM2、OpenVSLAM相比,其跟踪误差分别减少了68%和72%,验证了该方法在动态场景中位姿跟踪的准确性。     

基于语义一致性约束与局部-全局感知的多模态3D视觉定位

3D多模态数据稀缺,使得传统方法进行监督训练时文本与视觉特征缺乏语义一致性。同时传统方法还易忽视局部关系与全局信息,从而导致性能不佳。针对上述问题,提出了一种基于语义一致性约束与局部-全局感知的多模态3D视觉定位方法。首先,该方法通过蒸馏2D预训练视觉语言模型知识,帮助3D模型提取到点云-文本语义一致性特征;其次设计了局部-全局感知模块,不断补充增强候选目标特征,以更精确匹配目标。在现有的3D视觉定位数据集ScanRefer上进行的实验表明,该方法在Acc@0.25 IoU和Acc@0.5 IoU两个指标上分别达到了50.53%和37.67%,超越了现有大多数3D视觉定位算法,证实了该方法的有效性。     

基于多模态多级特征聚合网络的光场显著性目标检测

现有基于深度学习的显著性检测算法主要针对二维RGB图像设计,未能利用场景图像的三维视觉信息,而当前光场显著性检测方法则多数基于手工设计,特征表示能力不足,导致上述方法在各种挑战性自然场景图像上的检测效果不理想。提出一种基于卷积神经网络的多模态多级特征精炼与融合网络算法,利用光场图像丰富的视觉信息,实现面向四维光场图像的精准显著性检测。为充分挖掘三维视觉信息,设计2个并行的子网络分别处理全聚焦图像和深度图像。在此基础上,构建跨模态特征聚合模块实现对全聚焦图像、焦堆栈序列和深度图3个模态的跨模态多级视觉特征聚合,以更有效地突出场景中的显著性目标对象。在DUTLF-FS和HFUT-Lytro光场基准数据集上进行实验对比,结果表明,该算法在5个权威评估度量指标上均优于MOLF、AFNet、DMRA等主流显著性目标检测算法。     

The effect of text and background colour on visual search of Web pages

Usability and visual impact in Web pages are not necessarily compatible ideals. This paper investigates the effect of colour on the presentation of information in a navigation bar, and aims to contribute towards design guidelines for the use of colour on the Web. We studied the effect of the combination of text and background colour on visual search performance and subjective preference. Twenty-nine participants carried out a visual search task using mock Web pages. Analysis showed that higher contrasts between text and background colour led to faster searching and were rated more favourably. The results are discussed in terms of visual search processes and design recommendations are given.     

Lighting for road tunnels: The influence of CCT of light sources on reaction time

Drivers’ visual performance is closely related to traffic safety in a real driving environment. In order to improve the traffic safety of road tunnel lighting, the effect of LED lighting on human visual performance was investigated using reaction time as a parameter. The experiment was performed with a scale model that can simulate a road tunnel lighting environment. Reaction times were measured under different values of luminance, correlated colour temperature (CCT), eccentricity, and contrast. The results show that visual performance can be improved by increasing the CCT of the light sources. The improvement of visual performance is greater in peripheral vision than that in foveal vision. The shortest reaction times were measured at a luminance level of 10 cd/m² and at a CCT of 5000 K. An appropriate luminance value with high CCT is recommended for tunnel lighting in interior and transition zones.     

Integrating complex information with object displays: psychophysical evaluation of outlines

The interactive use of visual interface tools has diversified the use of visualisations. This article reviews the relevant aspects of interaction and challenges the sufficiency of traditional evaluation criteria developed for static graphs. Traditionally, the problem for statisticians has been to maintain perceptual discriminability of details, when quantities of data increase. Currently, however, even non-professional users need to integrate qualitatively different kinds of information. The review of task requirements indicates the use of a visual outline: (1) visual tools can facilitate parallel separation of individual data entities and integration of their features and (2) more focused comparisons require visual memory due to eye movements. The article reports psychophysical experiments that measure performance accuracy and response latency conditioned by the above task requirements. The impact of shape and colour on performance interacted with display times; the times were shorter (100 ms) or longer (1 s) than the duration of typical gaze fixation. The features of graphs in the experiments were derived from a popular internet service. Thus, we describe methods for evaluating visual components of real services and provide general guidelines for visual design of human–computer interaction.     

基于注意力感知和语义感知的RGB-D室内图像语义分割算法

近年来,全卷积神经网络有效提升了语义分割任务的准确率.然而,由于室内环境的复杂性,室内场景语义分割仍然是一个具有挑战性的问题.随着深度传感器的出现,人们开始考虑利用深度信息提升语义分割效果.以往的研究大多简单地使用等权值的拼接或求和操作来融合RGB特征和深度特征,未能充分利用RGB特征与深度特征之间的互补信息.本文提出...     

New electronic media and the human interface

Discussion is focused on the relationship between the new media and the human visual system. First, the process of information transmission in the human visual system is shown as a hierarchically structured model. Second, examples of the human interface in the development of a Japanese teletext system are shown, such as desired luminance and contrast, or favourable colour combination, or how to make superimposed sentences on screen easy to read. Third, examples of the human interface in the development of Hi-Vision, or High Definition Television (HDTV), are discussed, namely the conditions required for Hi-Vision from the standpoint of visual characteristics, the visual effects of Hi-Vision, and an objective evaluation of psychological effects of Hi-Vision.     

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.     

The effects of ambient illumination and target luminance on colour coding in a CRT display

In this paper, the effects of absolute colour-identification on a CRT display are compared under different experimental conditions, i.e., ambient illuminant intensity, colour temperature and target luminance. The results indicate that performance of colour identification deteriorated as ambient illuminant intensity was increased. Colour identification under the 2800 K colour temperature was better than that under the 1800 K and 5800K. Performance at high luminance was superior to that at low luminance. Colours were chosen to make up the colour coding system on the CRT display under various ambient illuminant conditions. The number and range of the colours selected for the colour coding could be expanded by increasing the target luminance on the CRT display.     

Effects of the viewing context on target detection. Implications for road lighting design

The Small Target Visibility (STV) model is the main model used to assess the quality of road lighting installations (IESNA, 2000). However, this model is based on a simple detection task in foveal vision using psychophysical data from laboratory conditions. The purpose of this study was to evaluate the impact of a complex background and apparent motion on target detection performance in mesopic vision, for three luminance contrasts, with reference to the STV scenario. To do so, participants were invited to detect standard square targets varying in terms of contrast presented in three Conditions: a uniform background, still images, and a video. Luminance levels were chosen in the mesopic domain relevant for road lighting at night. Images and video were chosen in relation to a driving task at night. The results showed that both the spatial context and the apparent motion had a negative impact on peripheral target detection performance: contrasts which are easy to detect in conditions close to the STV reference data may lead to poor performance if one adds context variables. These results give evidence that the STV model used for road lighting design based on laboratory data is limited, which strengthens previous results (Mayeur et al., 2008). The results are discussed in relation to the field factor used by practitioners to compensate for the differences between the STV reference scenario (detection of a small square target on a lit road while driving) and the STV psychophysical reference data.     

The influences of tunnel lighting environment on drivers’ peripheral visual performance during transient adaptation

Highway tunnel lighting environments include sidewall, pavement, ceiling, etc., their surface luminance were affected by the distribution of luminous flux of the lamps. While driving through a highway tunnel, the driver needs to complete a series of visual tasks. The spatial distribution of the luminance and color on the road surface and sidewall, as well as the correlated color temperature, influence the driver’s visual performance. The process of driving through a highway tunnel lighting environment was simulated in a laboratory where the luminance of the non-uniform visual environment was gradually decreased. The effects of the spatial distribution of the lighting environment parameters on a driver’s visual performance was studied by testing each subject’s reaction times and missed target rates. The tests showed that the spatial distribution of the lighting environment parameters significantly influences a driver’s visual performance, and the effects on test subjects’ peripheral visual performance are different in threshold, transition, and interior zones. The optimization of the spatial distribution of the lighting environment parameters can enhance traffic safety and energy saving of highway tunnel lighting.     

感知归类在目标识别中的应用研究

感知归类是指人们在不拥有任何图像内容知识的前提下，从底层原始图像特征中提取有效的图像联系的一种视觉能力。本文着重介绍了感知归类方法在目标识别中的应用，通过使用感知归类中的接近、相似、连续等概念，逐步将底层的图像特征组合为高层的图像联系，直至获得有意义的语义表征。     

基于模态预融合的三维指称表达理解

三维指称表达理解（3D VG）旨在通过理解指称表达来准确定位三维场景中的目标对象。现有3D VG研究通过引入文本和视觉分类任务优化文本和视觉编码器,这种方法可能由于文本和视觉特征的语义不对齐,从而导致模型难以在场景中定位文本描述的视觉对象。此外,3D VG数据集有限的数据量和复杂的模型结构往往导致模型过拟合。针对上述问题提出MP3DVG模型,通过学习统一的多模态特征表示完成单模态分类和3D VG任务,并降低模型的过拟合。基于跨模态特征交互提出TGV和VGT模块,在单模态任务之前预融合文本和视觉特征,减小不同模态特征因语义不对齐带来的不利影响。基于线性分类器可评价样本特征多样性的特性,提出周期性初始化的辅助分类器,并通过动态损失调节项自适应地调节样本损失,弱化模型的过拟合。大量实验结果表明所提方法的优越性,相比于MVT模型,MP3DVG在Nr3D和Sr3D数据集上性能分别提升1.1%和1.8%,模型的过拟合现象得到显著改善。     

Seamless,Static Multi‐Texturing of 3D Meshes

In the context of 3D reconstruction, we present a static multi‐texturing system yielding a seamless texture atlas calculated by combining the colour information from several photos from the same subject covering most of its surface. These pictures can be provided by shooting just one camera several times when reconstructing a static object, or a set of synchronized cameras, when dealing with a human or any other moving object. We suppress the colour seams due to image misalignments and irregular lighting conditions that multi‐texturing approaches typically suffer from, while minimizing the blurring effect introduced by colour blending techniques. Our system is robust enough to compensate for the almost inevitable inaccuracies of 3D meshes obtained with visual hull–based techniques: errors in silhouette segmentation, inherently bad handling of concavities, etc.     

Perceptually meaningful image editing

We introduce a novel concept called perceptually meaningful image editing and present techniques for manipulating the apparent depth of objects and creating the illusion of motion in 2D images. Our techniques combine principles of human visual perception with approaches developed by traditional artists. For our depth manipulation technique, the user loads an image, selects an object and specifies whether the object should appear closer or further away. The system automatically determines luminance or color temperature target values for the object and/or background that achieve the desired depth change. Our approach for creating the illusion of motion exploits the differences between our peripheral vision and our foveal vision by introducing spatial imprecision to the image.     

一种基于多层语义特征的图像理解方法

视觉场景理解包括检测和识别物体、推理被检测物体之间的视觉关系以及使用语句描述图像区域.为了实现对场景图像更全面、更准确的理解,将物体检测、视觉关系检测和图像描述视为场景理解中3种不同语义层次的视觉任务,提出一种基于多层语义特征的图像理解模型,并将这3种不同语义层进行相互连接以共同解决场景理解任务.该模型通过一个信息传递图将物体、关系短语和图像描述的语义特征同时进行迭代和更新,更新后的语义特征被用于分类物体和视觉关系、生成场景图和描述,并引入融合注意力机制以提升描述的准确性.在视觉基因组和COCO数据集上的实验结果表明,所提出的方法在场景图生成和图像描述任务上拥有比现有方法更好的性能.     

基于多级全局信息传递模型的视觉显著性检测

对神经网络中的卷积特征采用分层处理的思想能明显提升显著目标检测的性能。然而,在集成分层特征时,如何获得丰富的全局信息以及有效融合较高层特征空间的全局信息和底层细节信息仍是一个没有解决的问题。为此,提出了一种基于多级全局信息传递模型的显著性检测算法。为了提取丰富的多尺度全局信息,在较高层级引入了多尺度全局特征聚合模块（MGFAM）,并且将多层级提取出的全局信息进行特征融合操作;此外,为了同时获得高层特征空间的全局信息和丰富的底层细节信息,将提取到的有判别力的高级全局语义信息以特征传递的方式和较低层次特征进行融合。这些操作可以最大限度提取到高级全局语义信息,同时避免了这些信息在逐步传递到较低层时产生的损失。在ECSSD、PASCAL-S、SOD、HKU-IS等4个数据集上进行实验,实验结果表明,所提算法相较于较先进的NLDF模型,其F-measure（F）值分别提高了0.028、0.05、0.035和0.013,平均绝对误差（MAE）分别降低了0.023、0.03、0.023和0.007。同时,所提算法在准确率、召回率、F-measure值及MAE等指标上也优于几种经典的图像显著性检测方法。