光学学报, 2018, 38 (12): 1210002, 网络出版: 2019-05-10   

基于稀疏表示的失真卫星立体图像全参考质量评价 下载: 784次

Sparse Representation-Based Full-Reference Quality Assessment of Distorted Satellite Stereo Images
作者单位
宁波大学信息科学与工程学院, 浙江 宁波 315211
摘要
针对建筑物检测特定应用,提出一种基于稀疏表示的失真立体图像全参考质量评价方法。首先构建了一种新的失真卫星立体图像数据库,使用角点检测和数字表面模型的高程信息进行建筑物检测,并根据失真图像检测角点变化,提出检测准确率指标来表示图像的失真程度;然后提出一种基于稀疏表示的客观评价模型,其分别提取原始图像和失真图像的尺度不变特征转换和二进制稳健不变尺度特征进行字典学习;利用稀疏表示测量原始图像和失真图像之间的相似性,得到4个质量分数;最后通过支持向量回归融合4个质量分数得到最终的客观评价值。在构建的数据库上进行测试,实验结果表明,皮尔逊线性相关系数值高于0.90,斯皮尔曼等级相关系数值高于0.87,与现有的评价方法相比,所提方法能更好地反映卫星立体图像的质量。
Abstract
Aim

ing at the specific application of building detection, a sparse representation-based full-reference quality assessment method for distorted satellite stereoscopic images is proposed. First, a new distorted satellite stereo image database is constructed, in which the corner detection and the digital surface model elevation information are used for building detection. And a detection accuracy index is proposed to represent the degree of distortion based on the change of the detected corners. Then, an objective evaluation model based on sparse representation is proposed, which extracts scale-invariant feature transforms and binary robust invariant scalable key points of the original and the distorted images for dictionary learning. Four quality scores are obtained using sparse representation to measure the similarity between the original and the distorted images. Finally, the final objective assessment value is obtained by fusing the four quality scores using support vector regression. The test is carried out on the constructed database. The test results on the constructed database show that the Pearson linear correlation coefficient is higher than 0.90, and the Spearman rank correlation coefficient is higher than 0.87. Compared with the existing assessment methods, the proposed objective evaluation method can better reflect the quality of satellite stereo images.

熊义明, 邵枫, 孟祥超. 基于稀疏表示的失真卫星立体图像全参考质量评价[J]. 光学学报, 2018, 38(12): 1210002. Yiming Xiong, Feng Shao, Xiangchao Meng. Sparse Representation-Based Full-Reference Quality Assessment of Distorted Satellite Stereo Images[J]. Acta Optica Sinica, 2018, 38(12): 1210002.

本文已被 1 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!