摘要
为了解决黑白图像自动染色的难题,提出了一种基于Gabor小波的渐进式着色算法。该算法首先使用Gabor小波对黑白图像的纹理特征进行分析,在此基础上,根据纹理特征差异重新定义像素的邻居关系,最后利用最优化方法对染色问题进行迭代求解。该算法主要的创新点是交互操作少,并允许用户逐步添加色彩细节。同时该算法还是天然并行的,能够利用图形处理器(GPU)进行实时计算。为该算法和当今流行的着色算法做了效果对比,并且进行了效率分析,实验结果表明了该算法的可用性和效率。
To solve the problem of automatically colorizing gray images, a novel progressive method based on Gabor wavelets is presented.First the texture features of the image is analyzed by Gabor wavelet.Then the neighboring relationships are redefined according to the differences among texture features.Finally, the colorization problem is solved iteratively with an optimization method.The main contribution of this method is that only very few interactive operations are required and users are allowed to add the detailed color information step by step.And also, it is a natural parallelized algorithm and can be implemented in real-time with GPU.The effect and efficiency of the method are clearly demonstrated in the experiments.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第6期1327-1329,1334,共4页
Computer Engineering and Design
基金
国家自然科学基金项目(60573155)
关键词
着色
GABOR小波
最优化
并行
GPU
colorization
Gabor wavelet
optimization method
parallel
GPU
