【作者】 谢成军；

【导师】 檀结庆；

【作者基本信息】 合肥工业大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 数字图像修补，是利用受损区域周围的图像信息给受损区域填充信息的一门技术，本质上是一种图像插值问题。它属于图像复原的研究领域，是图像处理领域的一个分支。在多余物体的剔除(如视频图像和一些照片中删除部分人物、文字、小标题等)等方面得到了广泛的应用。除此之外，在博物馆中文物的保护方面(如对珍贵的书法与绘画作品进行的修复工作)也得到了很好的应用。正是图像修补技术的广泛应用，才使得这一技术成为了近几年的研究热点。本文首先比较全面的综述了图像修补中的研究现状，讨论了图像修补的特殊性以及方法学。然后深入研究了现行的几种图像局部性修补的数学模型和修补算法。实现了基于偏微分方程的BSCB模型、基于贝叶斯理论的整体变分修补模型，并对比了不同模型的图像修补效果。上述模型能够有效处理图像的几何特征，适应于非纹理图像的修补，但对富于纹理的图像修补效果不好。因此，本文给出了一种新的基于样本的图像修补模型，该模型通过计算样本填充的优先次序，将图像中的纹理与结构信息同时进行合理的填充，得到更令人满意的效果。本文最后给出了两种兼顾图像受损结构和纹理区域的自适应修补算法，将受损纹理区域与结构区域有效区分，然后分别进行处理，得出良好的图像修补结果。更多还原

【Abstract】 Digital image inpainting is to fill in image information on a blank domain based on the image information available outsidd. It is an interpolation problem, which belongs to the field of restoration and it’s also a branch of image processing. Applications of image inpainting range from restoration of photographs, films and paintings, to removal of occlusions, such as large unwanted regions, superimposed text, subtitles, stamps from images. In addition, it is more significant to restore the precious calligraphies and paintings in the museum with image inpainting technique. The widely-used applications of the image inpainting make this technique a hot research in the image processing techniques recently.Firstly,we summarize the recent work in this area ,discuss the particularity and methodology of inmage inpainting. Then, we are interested in the local inpainting methods, study several inpainting methods especially the Bertalmio- Sapiro-Caselles-B allester (BSCB) model and total variation (TV) model based on the Bayesian and variational principle with the comparision of results between two models.It is more effective for two models mentioned above to deal with the geometrical character of image and adaptive to non-texture. Therefore, in this thesis, we present a new image inpainting method based on the existing exemplar-based image inpainting idea. This method can obtain preferable results in human vision by rational Confidence and Data computing method.Finally, to overcome the disadvantage of those models mentioned above, we present two adaptive algorithms which combine image decomposition with local inpainting based on PDE and texture synthesis. Applying local inpainting and synthesis to non-texture and texture respectively gives satisfying results.更多还原