【摘要】 提出一种新的图像纹理替换方法,该方法在原始图像精确几何及光照信息未知的情况下,用新的纹理有效替换原始图像目标区域的纹理并逼真地保持了原始纹理的扭曲变形和光照效果.该方法借助反求的法向信息,通过一种基于网格边长伸展的参数化方法优化求得原始图像目标区域上划分的网格在纹理空间中的对应网格,有效模拟了原始纹理由于隐藏几何和透视投影引起的纹理扭曲效果;进而通过YCbCr颜色空间转换和传递保持了原始纹理的光照阴影效果.由于求取纹理采样规则的优化方法仅仅与原始图像空间相关,而与替换纹理无关,因而对目标图像处理一次后,可以有效地应用于任意新纹理的替换.通过适当的交互,替换纹理的纹元尺度渐变及纹理自遮挡效果亦可有效模拟.实验展示了良好的纹理替换效果.更多还原

【Abstract】 This paper presents a new method for replacing the existing textures in the interest region of an image with the new ones, without knowing the accurate underlying geometry and lighting conditions. A stretch-based parametrization incorporating the recovered normal information is deduced to imitate the texture distortion induced by the underlying geometry as well as the perspective distortion. The luminance of the original image is simultaneously preserved through color transfer on YCbCr color space. The method is independent on the replaced textures, however only correlates with the original image space. So after processing the input image only once, any new textures can be replaced efficiently without re-execution. By means of the parametrization, the replaced effects with progressively-variant texton scales and texture discontinuity in self-occlusion regions can be produced conveniently and realistically. Experimental results demonstrate satisfactory texture replacing results.更多还原