Literature visualization has particular characteristics:multimodality in production;multimediality in the means of distribution;visualization of language forms.The basic changes of literature lie in mode- choosing ori...Literature visualization has particular characteristics:multimodality in production;multimediality in the means of distribution;visualization of language forms.The basic changes of literature lie in mode- choosing orientation and potentials that language and image offer for making meaning.Under the framework of multimodal semiotics,image and language are two modes nestled among a multimodal ensemble.Each mode and the interactions between them have the potential to contribute equally to meaning.Elucidating "image-language relation" from multimodality can make it clear that the particular characteristics of image and language don't change;the former describes visual effect,the latter constructs image with imagination.展开更多
光场图像作为一种能够捕获场景每个位置光线信息的图像类型,在电子成像、医学影像和虚拟现实等领域具有广泛的应用前景.光场图像质量评估(Light Field Image Quality Assessment,LFIQA)旨在衡量此类图像的质量,但当前方法面临视觉效果...光场图像作为一种能够捕获场景每个位置光线信息的图像类型,在电子成像、医学影像和虚拟现实等领域具有广泛的应用前景.光场图像质量评估(Light Field Image Quality Assessment,LFIQA)旨在衡量此类图像的质量,但当前方法面临视觉效果与文本模态间异构性的重要挑战.为解决上述问题,本文提出了一种基于文本-视觉的多模态光场图像质量评估模型.具体来说,在视觉模态方面,我们设计了多任务模型,结合边缘自动阈值算法有效丰富了光场图像的关键表示特征.在文本模态方面,基于输入噪声特征与预测噪声特征的对比,准确识别光场图像的噪声类别,并验证了噪声预测对优化视觉表示的重要性.基于上述研究,进一步提出了一种优化的通用噪声文本配置方法,并结合边缘增强策略,显著提升了基线模型在光场图像质量评估中的准确性和泛化能力.此外,通过消融实验,评估了各组件对整体模型性能的贡献,验证了本文方法的有效性和稳健性.实验结果表明,该方法不仅在公开数据集Win5-LID和NBU-LF1.0的实验中表现出色,还在融合数据集中展示出优秀的实验结果,与现有最优算法相比,本文所提方法在两个数据库中的性能分别提升了2%和6%.本文提出的噪声验证策略和配置方法不仅为图像质量评估中的噪声预测任务提供了有价值的参考,也可用于其它噪声预测类型的辅助任务.展开更多
基金funded by 2013 humanistic and social science research project of the Education Department of Anhui Province(2013AJRW0150)2014 Planning Project of Philosophy and Social Sciences of Anhui Province(AHSKY2014D140)+1 种基金Project No.(JS2015HGXJ0074)Supported by the Fundamental Research Funds for the Central Universities
文摘Literature visualization has particular characteristics:multimodality in production;multimediality in the means of distribution;visualization of language forms.The basic changes of literature lie in mode- choosing orientation and potentials that language and image offer for making meaning.Under the framework of multimodal semiotics,image and language are two modes nestled among a multimodal ensemble.Each mode and the interactions between them have the potential to contribute equally to meaning.Elucidating "image-language relation" from multimodality can make it clear that the particular characteristics of image and language don't change;the former describes visual effect,the latter constructs image with imagination.
文摘光场图像作为一种能够捕获场景每个位置光线信息的图像类型,在电子成像、医学影像和虚拟现实等领域具有广泛的应用前景.光场图像质量评估(Light Field Image Quality Assessment,LFIQA)旨在衡量此类图像的质量,但当前方法面临视觉效果与文本模态间异构性的重要挑战.为解决上述问题,本文提出了一种基于文本-视觉的多模态光场图像质量评估模型.具体来说,在视觉模态方面,我们设计了多任务模型,结合边缘自动阈值算法有效丰富了光场图像的关键表示特征.在文本模态方面,基于输入噪声特征与预测噪声特征的对比,准确识别光场图像的噪声类别,并验证了噪声预测对优化视觉表示的重要性.基于上述研究,进一步提出了一种优化的通用噪声文本配置方法,并结合边缘增强策略,显著提升了基线模型在光场图像质量评估中的准确性和泛化能力.此外,通过消融实验,评估了各组件对整体模型性能的贡献,验证了本文方法的有效性和稳健性.实验结果表明,该方法不仅在公开数据集Win5-LID和NBU-LF1.0的实验中表现出色,还在融合数据集中展示出优秀的实验结果,与现有最优算法相比,本文所提方法在两个数据库中的性能分别提升了2%和6%.本文提出的噪声验证策略和配置方法不仅为图像质量评估中的噪声预测任务提供了有价值的参考,也可用于其它噪声预测类型的辅助任务.
