Stereoscopic view synthesis based on region-wise rendering and sparse representation

Depth image-based rendering (DIBR), which is used to render virtual views with a color image and the corresponding depth map, is one of the key techniques in the 2D to 3D conversion process. One of the main problems in DIBR is how to reduce holes that occur on the generated virtual view images. In this paper, we make two main contributions to deal with the problem. Firstly, a region-wise rendering framework, which divides the original image regions into three special classes and renders each with optimal adaptive process respectively, is introduced. Then, a novel sparse representation-based inpainting method, which can yield visually satisfactory results with less computational complexity for high quality 2D to 3D conversion, is proposed. Numerical experimental results demonstrate the good performance of the proposed methods.     

基于多重稀疏字典的行人检测方法

针对于稀疏编码在行人检测问题中提取的特征维数高和不能够有效描述行人的问题,提出了一种基于多重稀疏字典直方图的特征提取方法。通过稀疏表示方法,预先学习多个不同稀疏度的字典,分别利用每一个字典对行人图像进行稀疏编码,统计每个字典中对应稀疏编码单元的分布直方图作为行人图像的特征描述子。该方法提取到的特征维数低,并且能够有效地描述行人,具有良好的检测性能。     

稀疏阈值的超分辨率图像重建

为了解决基于字典学习的超分辨重构算法耗时过长的问题,提出了基于稀疏阈值模型的图像超分辨率重建方法。首先,将联合字典理论与图像块稀疏阈值方法相结合,训练得到高、低分辨率过完备图像字典对。接着,通过稀疏阈值OMP算法对图像特征块进行稀疏表示。然后,通过高分辨率字典重构出初始的超分辨图像。最后,通过改进迭代反投影算法对初始的超分辨图像进行全局优化,从而进一步提高图像重构质量。实验结果表明,超分辨图像重构平均峰值信噪比(PSNR)为30.1 d B,平均结构自相似度(SSIM)为0.937 9,平均计算时间为10.2 s。有效提高了超分辨重构的速度,改善了重构高分辨图像的质量。     

A Method of Domain Dictionary Construction for Electric Vehicles Disassembly

Currently, there is no domain dictionary in the field of electric vehicles disassembly and other domain dictionary construction algorithms do not accurately extract terminology from disassembly text, because the terminology is complex and variable. Herein, the construction of a domain dictionary for the disassembly of electric vehicles is a research work that has important research significance. Extracting high-quality keywords from text and categorizing them widely uses information mining, which is the basis of named entity recognition, relation extraction, knowledge questions and answers and other disassembly domain information recognition and extraction. In this paper, we propose a supervised learning dictionary construction algorithm based on multi-dimensional features that combines different features of extraction candidate keywords from the text of each scientific study. Keywords recognition is regarded as a binary classification problem using the LightGBM model to filter each keyword, and then expand the domain dictionary based on the pointwise mutual information value between keywords and its category. Here, we make use of Chinese disassembly manuals, patents and papers in order to establish a general corpus about the disassembly information and then use our model to mine the disassembly parts, disassembly tools, disassembly methods, disassembly process, and other categories of disassembly keywords. The experiment evidenced that our algorithms can significantly improve extraction and category performance better than traditional algorithms in the disassembly domain. We also investigated the performance algorithms and attempts to describe them. Our work sets a benchmark for domain dictionary construction in the field of disassembly of electric vehicles that is based on the newly developed dataset using a multi-class terminology classification.     

游客情感计算的文本大数据挖掘方法比较研究

旅游文本大数据以其方便、快捷和低门槛的特点为游客情感计算提供了极大便利,已经成为旅游大数据的主要来源之一。基于大数据理论和情感理论,以文本大数据为数据源,在全面梳理国内外情感计算相关成果的基础上,利用人工智能中的逻辑/算法编程方法、机器学习方法、深度学习方法对旅游文本大数据进行挖掘,探索最佳的基于文本大数据的游客情感计算方法。研究发现：（1）基于情感词典的游客情感计算模型,其核心是构建情感词典和设计情感计算规则,方法简单,容易实现,适用语料范围广。（2）机器学习,用统计学方法抽取文本中的特征项,具有非线性特征,可靠性较线性特征的情感词典方法高。（3）基于深度学习技术的游客情感计算,效果良好,准确率在85%以上。训练多领域的文本语料易于移植,实用性强,且泛化能力好,较适合大数据时代游客情感计算研究。     

Nonparametric Bayesian dictionary learning algorithm based on structural similarity

Though nonparametric Bayesian methods possesses significant superiority with respect to traditional comprehensive dictionary learning methods,there is room for improvement of this method as it needs more consideration over the structural similarity and variability of images.To solve this problem,a nonparametric Bayesian dictionary learning algorithm based on structural similarity was proposed.The algorithm improved the structural representing ability of dictionaries by clustering images according to their non-local structural similarity and introducing block structure into sparse representing of images.Denoising and compressed sensing experiments showed that the proposed algorithm performs better than several current popular unsupervised dictionary learning algorithms.     

基于Gabor特征和字典学习的高斯混合稀疏表示图像识别

为了克服图像识别中光照,姿态等变化带来的识别困难,同时提高稀疏表示图像识别的鲁棒性,本文提出了一种基于Gabor特征和字典学习的高斯混合稀疏表示图像识别算法.高斯混合稀疏表示是基于最大似然估计准则,将稀疏保真度表示为余项的最大似然函数,最终识别问题转化为求解加权范数的优化逼近问题.本文算法首先提取图像的Gabor特征;然后对Gabor特征集进行字典学习,由于在学习过程中引入了Fisher准则作为约束,学习得到具有类别标签的新字典;最后使用高斯混合稀疏表示识别方法进行分类识别.在3个公开数据库(人脸数据库AR库和FERET库以及USPS手写数字库)上的实验结果验证了该算法的有效性和鲁棒性.     

抗年龄干扰的人脸识别

本文将人脸图像特征分解为身份特征部分和年龄干扰特征部分,并分别投影到两个独立的子空间.为了提高由字典张成的特征子空间的表征能力和区分能力,字典学习过程中同时引入了人脸图像的重构误差约束项和类别约束项.因此,任意的人脸特征都可以由学习到的身份字典和年龄字典投影到对应的身份子空间和年龄子空间,然后再基于身份子空间进行人脸识别,从而使年龄的干扰得到了有效的抑制.通过在MORPH和FGNET数据库上的实验,证实了基于年龄不变的身份特征子空间学习方法能提升人脸识别性能.     

基于K-SVD字典稀疏分解的实木地板去噪方法

针对实木地板的图像获取过程中,所产生的噪声问题,引入了K-SVD字典的学习算法,提出了一种图像的有用信息稀疏分解去噪的方法,目的是有效的保留实木地板的有用纹理信息,并抑制其中掺杂的噪声。通过对图像稀疏分解后得到的值,来进行图像重构,就可以达到图像的去噪目的。首先,构造一个初始化的DCT字典,对图像分块处理;接着,在这个初始化字典的基础之上,进行纹理信息的稀疏分解,同时,对它们之间的残差值进行奇异值分解,更新字典;最后,利用得出的最优化字典,采用正交匹配重构算法,完成去噪图像的重建。实验表明,该算法得出的图像主观效果好,减少了去噪后的模糊程度及保留更多细节信息,在不同程度的噪声下,PSNR较高。     

基于噪声检测的置换混叠图像盲分离

置换混叠图像盲分离作为一类新型的单信道盲分离,在理论和方法上与传统的单信道盲分离有着本质的不同。针对一类置换区域含噪声的置换混叠图像,本文提出一种基于噪声检测的置换混叠图像盲分离算法。本文首先给出置换混叠图像的数学模型,对置换混叠图像用非零元个数约束的K-SVD算法进行训练得到其稀疏表示的字典,利用学习得到的字典对置换混叠图像去噪,然后利用去噪后的置换混叠图像与原图像作差运算得到差图像,通过检测差图像来确定出置换区域的位置和大小。并利用图像形态学运算优化置换区域,采用阈值化操作分离出置换图像。实验结果表明,本文算法能够较好的从置换混叠图像中分离出置换图像,并且不受置换图像的大小、位置、个数和置换图像所含噪声大小的影响。