作　　者： ; ;

机构地区： 韶关大学计算机科学学院计算机科学与技术系

出　　处： 《信息与电子工程》 2006年第6期440-443,共4页

摘　　要： 在各种超分辨率图像重构算法中,最大后验概率(Maximum a Posteriori,MAP)算法因其具有优异的重构性能而受到广泛关注。但由于目前在MAP算法中普遍采用的是平滑型图像先验模型,导致重构出来的图像边界不明锐,一些细节不清晰。本文提出了一种新的边界增强型图像先验模型。不同于已有的图像模型,新模型对图像中的非连续性不是进行惩罚,而是进行增强。实验结果表明,新模型能够获得优于平滑型图像先验模型的重构效果。 The Super-Resolution （SR） technique has been proven to be extremely useful in satellite imaging, video surveillance, medical computed tomographic imaging, and other applications. Among all the published SR techniques, Maximum a Posteriori（MAP） algorithm has received much attention for its outstanding reconstruction quality. However, due to the use of smooth prior image model in MAP, the reconstructed edges and some image details are blurred. In this paper, a new edge-enhancing prior image model is proposed to solve this problem. Unlike the previously published models, the new model does not punish, but enhance the discontinuities in the image. Experiment results demonstrate that the new model can provide better reconstruction quality than the smooth prior image models.

关 键 词： 超分辨率 最大后验概率 图像先验模型 图像重构

领　　域： [电子电信] [电子电信]