سال انتشار: ۱۳۸۹
محل انتشار: ششمین کنفرانس ماشین بینایی و پردازش تصویر
تعداد صفحات: ۶
Seyed Mahdi Shariatzadeh – Sharif University of Technology, Faculty of Mathematical Sciences, Computer Science Department, Tehran
Shohreh Kasaei – Sharif University of Technology, Department of Computer Engineering, Tehran
Nezam Mahdavi-Amiri – Sharif University of Technology, Faculty of Mathematical Sciences, Computer Science Department
This paper presents a multi-resolution approach for completion and object removal of digital images using valid patches of the image. While the multi-resolution hierarchy provides a good ability to complete large holes in different resolutions, a priority-based best-first algorithm is used to refine the result at each level. It tries to maintain the content structure when estimating the texture inside the missing area. Also, in order to improve the quality of patch search, different patch sizes are employed for search and completion. Demonstrated experimental results show the efficiency of the proposed algorithm for a variety of input images in comparison with a number of recent works, but with relatively reduced computational complexity.