Efficient Texture Creation Based on Random Patches in Database and Guided Filter

Seok Bong YOO  Mikyong HAN  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E101-D   No.11   pp.2840-2843
Publication Date: 2018/11/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.2018EDL8143
Type of Manuscript: LETTER
Category: Image Processing and Video Processing
Keyword: 
texture creation,  image super-resolution,  guided filter,  random patches,  

Full Text: PDF(1.7MB)
>>Buy this Article


Summary: 
As the display resolution increases, an effective image upscaling technique is required for recent displays such as an ultra-high-definition display. Even though various image super-resolution algorithms have been developed for the image upscaling, they still do not provide the excellent performance in the ultra-high-definition display. This is because the texture creation capability in the algorithms is not sufficient. Hence, this paper proposes an efficient texture creation algorithm for enhancing the texture super-resolution performance. For the texture creation, we build a database with random patches in the off-line processing and we then synthesize fine textures by employing guided filter in the on-line real-time processing, based on the database. Experimental results show that the proposed texture creation algorithm provides sharper and finer textures compared with the existing state-of-the-art algorithms.