CAJ | 학술논문

为了解决单幅图像超分辨重建的问题，该文提出了一种在稀疏表示理论框架下基于多成分字典的方法。首先根据图像的退化模型，深入分析高低分辨率图像的关系，得到高分辨图像可由低分辨率图像在对应字典下的稀疏系数来重构的结论。根据这一结论，采用多成分字典分别表示图像的不同结构特征，并用匹配追踪的方法得到低分辨图像在多成分字典下的表示系数，然后在对应的高分辨字典下对高分辨率的图像进行重建，实现了基于多成分字典的单幅图像超分辨率重建。该文所提的方法对单幅图像的超分辨率重建具有较好的通用性，相对于传统的超分辨重建算法，在自然图像和卡通图像的实验中验证了算法的有效性。
위료해결단폭도상초분변중건적문제，해문제출료일충재희소표시이론광가하기우다성분자전적방법。수선근거도상적퇴화모형，심입분석고저분변솔도상적관계，득도고분변도상가유저분변솔도상재대응자전하적희소계수래중구적결론。근거저일결론，채용다성분자전분별표시도상적불동결구특정，병용필배추종적방법득도저분변도상재다성분자전하적표시계수，연후재대응적고분변자전하대고분변솔적도상진행중건，실현료기우다성분자전적단폭도상초분변솔중건。해문소제적방법대단폭도상적초분변솔중건구유교호적통용성，상대우전통적초분변중건산법，재자연도상화잡통도상적실험중험증료산법적유효성。
In order to solve the problem of super-resolution reconstruction of single image,a hybrid approach is presented under the framework of sparse representation with multi-component dictionary. According to the image degradation model,the algorithms focus on the relationship between the low-resolution images and high-resolution images. This paper concludes that the high-resolution images can be reconstructed by the coefficients of low-resolution images in the corresponding dictionary. The sparse coefficients are obtained by the method of match pursuit based on the multi-component dictionary which indicates different structural characteristics of the image. The high-resolution images are reconstructed in the corresponding high-resolution dictionary. This paper introduces an objective and new strategy capable of efficiently guiding the image restoration. Extensive experimental studies conducted on the nature and cartoon images show the effectiveness of the proposed method.