CAJ | 학술논문

提出了一种新的基于联合字典学习的图像去噪方法。考虑图像复原问题总是涉及高、低质量两个版本的图像，从大量图像的两个版本中成对采样联合训练字典。所得字典不仅具有某类样本图像的结构特征，更具有一般图像这两个版本之间的对应关系，因此使得图像复原估计更有指向性，复原结果与原图像在细节上也更接近。
제출료일충신적기우연합자전학습적도상거조방법。고필도상복원문제총시섭급고、저질량량개판본적도상，종대량도상적량개판본중성대채양연합훈련자전。소득자전불부구유모류양본도상적결구특정，경구유일반도상저량개판본지간적대응관계，인차사득도상복원고계경유지향성，복원결과여원도상재세절상야경접근。
A novel approach of denoising is presented based on joint-dictionary learning .It is considered that there are always two versions of images with high and low quality in image restoration problems .The samples are taken from a pair of these images once a time for joint-dictionary training .As a result , a dictionary containing not only the samples'structure features but also the relationship between images of these two versions is obtained .So when a degraded image is to be recovered , the joint-learned dictionary will work as an assistance to help to get a more precise estimation to the original image .