Image stylization with enhanced structure on GPU  

作　　者：LI Ping SUN HanQiu SHENG Bin SHEN JianBing 

机构地区：[1]Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, China [2]Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China [3]School of Computer Science and Technology, Beijing Institute of Technology, Beijing 100081, China

出　　处：《Science China(Information Sciences)》2012年第5期1093-1105,共13页中国科学（信息科学）（英文版）

基　　金：supported by an RGC research grant (ref. 416007);UGC direct grants for research (Grant No.2050454,2050485);National Natural Science Foundation of China (Grant No. 60903068);Key Program of NSFC-Guangdong Union Foundation (Grant No. U1035004)

摘　　要：This paper presents a graphics processing unit （GPU） based stylization approach that preserves the fine structure between the original and the stylized images using gradient optimization.Existing abstraction and painterly stylization methods focused on contrast manipulation only,and thus the detailed salient structures of the input images are always destroyed when performing the current stylization techniques because of limitations like unavoidable salience information loss caused by contrast abstraction.We propose an image structure map to naturally model the fine structure existing in the original images.Gradient-based structure tangent generation and tangent-guided image morphology are used to construct the structure map.The image structure map,unlike an edge map,not only systematically models the boundary information within the imagery but also accentuates the underlying inner structure detail for further stylization.We facilitate the final stylization via parallel bilateral grid and structure-aware stylizing optimization on a GPU-CUDA platform in real time.In multiple experiments,the proposed method consistently demonstrates efficient and high quality image stylization performance.This paper presents a graphics processing unit （GPU） based stylization approach that preserves the fine structure between the original and the stylized images using gradient optimization.Existing abstraction and painterly stylization methods focused on contrast manipulation only,and thus the detailed salient structures of the input images are always destroyed when performing the current stylization techniques because of limitations like unavoidable salience information loss caused by contrast abstraction.We propose an image structure map to naturally model the fine structure existing in the original images.Gradient-based structure tangent generation and tangent-guided image morphology are used to construct the structure map.The image structure map,unlike an edge map,not only systematically models the boundary information within the imagery but also accentuates the underlying inner structure detail for further stylization.We facilitate the final stylization via parallel bilateral grid and structure-aware stylizing optimization on a GPU-CUDA platform in real time.In multiple experiments,the proposed method consistently demonstrates efficient and high quality image stylization performance.

关 键 词：image stylization non-photorealistic rendering real-time processing GPU image abstraction 

分 类 号：TP334.7[自动化与计算机技术—计算机系统结构] TS106.6[自动化与计算机技术—计算机科学与技术]