Evaluation of image similarity by histogram intersection

Colour is the most widely used attribute in image retrieval and object recognition. A technique known as histogram intersection has been widely studied and is considered to be effective for color‐image indexing. The key issue of this algorithm is the selection of an appropriate color space and optimal quantization of the selected color space. The goal of this article is to measure the model performance in predicting human judgment in similarity measurement for various images, to explore the capability of the model with a wide set of color spaces, and to find the optimal quantization of the selected color spaces. Six color spaces and twelve quantization levels are involved in evaluating the performance of histogram intersection. The categorical judgment and rank order experiments were conducted to measure image similarity. The CIELAB color space was found to perform at least as good as or better than the other color spaces tested, and the ability to predict image similarity increased with the number of bins used in the histograms, for up to 512 bins (8 per channel). With more than 512 bins, further improvement was negligible for the image datasets used in this study. © 2005 Wiley Periodicals, Inc. Col Res Appl, 30, 265–274, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.20122     

二元光栅照明研究

采用部分相干成象理论对二元光机照明进行了较为详细的研究。分析了二元光栅照明对分辨力，焦，照明能量和照明均匀性的影响，并通过实验研究验证了二元光栅对分辨力和焦深有较大程度的提高，同时能量损失不大。     

i－line光刻高均匀照明系统光能分布的计算模拟

叙述了ｉ－ｌｉｎｅ光刻高均匀照明系统的均匀照明原理，以及其具有特殊面形（非球面）、特殊结构（非共轴）、超光学表面（由多个非共轴光学表面组成）的光学系统的光能分布计算模拟的原理、方法。作为例子，用所编写的程序计算模拟了我们近来设计的一曝光系统几个重要光学表面上具有权重的点列图和光能分布曲线。     

Fall detection for multiple pedestrians using depth image processing technique

A fall detection method based on depth image analysis is proposed in this paper. As different from the conventional methods, if the pedestrians are partially overlapped or partially occluded, the proposed method is still able to detect fall events and has the following advantages: (1) single or multiple pedestrian detection; (2) recognition of human and non-human objects; (3) compensation for illumination, which is applicable in scenarios using indoor light sources of different colors; (4) using the central line of a human silhouette to obtain the pedestrian tilt angle; and (5) avoiding misrecognition of a squat or stoop as a fall. According to the experimental results, the precision of the proposed fall detection method is 94.31% and the recall is 85.57%. The proposed method is verified to be robust and specifically suitable for applying in family homes, corridors and other public places.     

Rough membership function based illumination classifier for illumination invariant face recognition

This paper proposes a face recognition system to overcome the problem due to illumination variation. The propose system first classifies the image's illumination into dark, normal or shadow and then based on the illumination type; an appropriate technique is applied for illumination normalization. Propose system ensures that there is no loss of features from the image due to a proper selection of illumination normalization technique for illumination compensation. Moreover, it also saves the processing time for illumination normalization process when an image is classified as normal. This makes the approach computationally efficient. Rough Set Theory is used to build rmf illumination classifier for illumination classification. The results obtained as high as 96% in terms of accuracy of correct classification of images as dark, normal or shadow.     

基于相频特性的具有光照鲁棒的人脸识别研究

传统人脸识别对人脸图像光照要求与人脸训练库差异不大,这提高了人脸识别系统运行的环境条件,限制了应用范围.为了降低人脸识别对环境条件的要求,克服光照对人脸识别的影响,分析了人脸图像的相频特性与光照的无关性,提出了基于相频特性的人脸识别,保留了人脸之间的可区分性.由于人脸之间可区分的信息量较少,运用最小非零特征向量作为人脸特征.实验仿真表明,与传统方法相比,提出的基于相频特性的人脸识别对光照变化具有鲁棒性.     

双摄像机下人体遮挡时的跟踪方法

针对双摄像机下存在人体遮挡情况时的跟踪问题,提出了利用人体3维位置信息来实现跟踪的方法。该方法首先对其中一个摄像机视频图像中的人体像素抽样,接着在其他摄像机视频图像中找出抽样像素的匹配点,计算出每一对匹配点在世界坐标系中所对应的3维点,然后依据3维位置信息将3维点聚类,找出每一个聚类区域中的3维点所对应的图像中的一组像素点,并对其构建高斯平滑直方图模型。在此基础上,依据直方图模型将互相遮挡的人体分割开来,最后根据求取的人体像素点的匹配关系来确定不同摄像机中同一个人的对应关系。实验结果表明,该方法能有效实现遮挡情况下的人体跟踪。     

Stereo Light Probe

In this paper we present a practical, simple and robust method to acquire the spatially‐varying illumination of a real‐world scene. The basic idea of the proposed method is to acquire the radiance distribution of the scene using high‐dynamic range images of two reflective balls. The use of two light probes instead of a single one allows to estimate, not only the direction and intensity of the light sources, but also the actual position in space of the light sources. To robustly achieve this goal we first rectify the two input spherical images, then, using a region‐based stereo matching algorithm, we establish correspondences and compute the position of each light. The radiance distribution so obtained can be used for augmented reality applications, photo‐realistic rendering and accurate reflectance properties estimation. The accuracy and the effectiveness of the method have been tested by measuring the computed light position and rendering synthetic version of a real object in the same scene. The comparison with standard method that uses a simple spherical lighting environment is also shown.     

面向不同数据分布的多维直方图算法COCA-Hist

基于代价的RDBMS优化器需要对含有范围查询的合取谓词的结果集基数进行准确的估计,多维直方图对多维数据分布进行模拟,避免在估计结果集基数时采用数据独立性假设,造成估计误差过大,进而导致选择非优化的查询执行计划.在不同的数据分布情况下,传统的多维直方图(如MHist-2)效果有很大不同.数据相关系数和值域密度、值域参数是准确刻画多维数据分布的有效指标,文中提出了根据不同的指标采用不同的动态优化的多维直方图算法COCA-Hist,可以大大改善传统多维直方图在平均情况下的准确性.通过分析传统的多维直方图的最坏情况,COCA-Hist的改进算法可以改进传统的多维直方图在最坏情况下的准确性.实验比较了COCA-Hist和传统的多维直方图MHist-2以及GENHist和STHoles的准确性和时间效率.实验显示无论在平均情况下还是在最坏情况下COCA-Hist的改进算法均优于传统的MHist-2直方图,并且COCA-Hist的准确性和创建时间均比GENHist有极大的改善,在准确性方面COCA-Hist较优于STHoles,而在空间预算有限时STHoles的创建时间比COCA-Hist高两个数量级.     

Estimating aerodynamic resistance of rough surfaces using angular reflectance

Current wind erosion and dust emission models neglect the heterogeneous nature of surface roughness and its geometric anisotropic effect on aerodynamic resistance, and over-estimate the erodible area by assuming it is not covered by roughness elements. We address these shortfalls with a new model which estimates aerodynamic roughness length (z₀) using angular reflectance of a rough surface. The new model is proportional to the frontal area index, directional, and represents the geometric anisotropy of z₀. The model explained most of the variation in two sets of wind tunnel measurements of aerodynamic roughness lengths (z₀). Field estimates of z₀ for varying wind directions were similar to predictions made by the new model. The model was used to estimate the erodible area exposed to abrasion by saltating particles. Vertically integrated horizontal flux (F_h) was calculated using the area not covered by non-erodible hemispheres; the approach embodied in dust emission models. Under the same model conditions, F_h estimated using the new model was up to 85% smaller than that using the conventional area not covered. These F_h simulations imply that wind erosion and dust emission models without geometric anisotropic sheltering of the surface, may considerably over-estimate F_h and hence the amount of dust emission. The new model provides a straightforward method to estimate aerodynamic resistance with the potential to improve the accuracy of wind erosion and dust emission models, a measure that can be retrieved using bi-directional reflectance models from angular satellite sensors, and an alternative to notoriously unreliable field estimates of z₀ and their extrapolations across landform scales.