摘要
引起图像退化的因素众多,因而难以用一个统一的方法来进行恢复处理。鉴于图像的像素和各颜色分量通道间本质上存在某种相关性,以及以支持向量机为核心的统计学习理论具有较好地解决小样本、非线性、高维数问题的能力,提出了一种新的空域图像恢复方法,并通过对来自于待处理图像本身的训练样本的学习,构造自适应的回归插值函数;然后基于该函数对图像作有选择的修改,从而达到图像恢复的目的。实验表明,该方法是有效的,并且具有较好的泛化性能。
Due to a variety factors causing image degradation, a unified method to restore image does not exist. As for some essential correlations between pixels and color channels in image, and the statistical theory based on the support vector machine being able to solve problems such as small sample, non-linear and high dimension, a novel spatial restoration approach is put forward. Firstly a learning procedure is performed on samples from the image to be processed, then a regression interpolation function is constructed, and the image restoration is implemented via modifying original image pix values based on the regression interpolation function. The experiments validate the efficiency and adaptability of the restoration approach.
出处
《海军工程大学学报》
CAS
北大核心
2007年第1期90-92,104,共4页
Journal of Naval University of Engineering
关键词
图像恢复
支持向量机
回归插值函数
空间分布相关性
image restoration
support vector machine
regression interpolation function
spatial distribution correlation
