摘要
超分辨率图像重建是一个不适定问题,学术上富有挑战性,在影像处理、高分辨率对地观测等领域具有广泛的用途,其目标评价特性与格式塔理论关于视像认知的相以性、共性线、同趋性等高度吻合.本文引入格式塔理论,在国际图像与视频压缩标准JPEG2000、MPEG-4推荐的提升小波分解与重建框架下,开展重建级大于分解级的超分辨率图像重建模型与算法研究,有如下三方面的贡献:1)对高频信息进行高精度估计,重建获得超分辨率图像,最高频系数置零的小波变换去噪为超分辨率图像重建提供了一个逆向成功的范例;2)遵循格式塔理论,对边缘轮廓、纹理及高频细节信息进行重建;3)建立了带格式塔约束的超分辨率图像重建优化模型框架,体现了人类视觉评价指标.
Super-resolution（SR） image reconstruction is an ill-posed which has been widely used in some challenging fields such as image processing and high-resolution for earth observation.The evaluation criteria of SR image reconstruction in terms of Video recognition like vicinity,similarity,and continuity of direction,are identical with the ones of Gestalt theory.Using the decomposition and reconstruction techniques of lifting wavelet recommended by JPEG2000 and MPEG-4,this paper applies Gestalt theory to study the model of SR image reconstruction in which the level of reconstruction is higher than the level of decomposition.Our main contributions include：1） Estimate the high frequency information with high precision,reconstruct SR image and provide a successful reverse example of SR image reconstruction which adopts wavelet denoising and sets the coefficient of the highest frequency as zero;2） Reconstruct the edge contour,texture and high frequency information according to Gestalt theory;3） Set up an optimal model for SR image reconstruction with the Gestalt constraints,which reflects the evaluation criteria of human vision.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2011年第2期261-270,共10页
Journal of Xiamen University：Natural Science
基金
国家重点基础研究发展计划(973)项目(2007CB311005)
国防基础科研计划项目(B1420110155)
福建省自然科学基金项目(A0710020)
中央高校基本科研业务费专项资金(2010121066
2010121067)
关键词
格式塔理论
超分辨率
图像重建
低可观测目标
正则性约束
Gestalt theory
super-resolution
image reconstruction
low observable target
regularization restriction
