光照条件变化下的图像生成方法

Image Generation on Illumination Changing

提出了两种图像生成的方法:(1)由图像分割、三维重建和投影生成构成.该方法考虑了图像中各部分表面发射模型的差异,通过将图像各区域划分成朗伯表面和镜面反射,分别进行三维重建,然后融合两结果,改变光照方向和强度,投影生成新的图像;(2)将Shape from shading和Shape from texture技术融合起来.采用Gabor滤波器将图像中的纹理成分和阴影成分区分开来,再用两种方法各自生成三维立体图像,依据它们的特性在频域融合两个三维图像,然后再改变视角和光照的强度、方向,生成仿真的二维图像.实验表明,由该两种方法恢复出的形状优于传统的估计方法,生成的图像真实感强.

This paper presented two kinds of image generation methods. The first method consists of image segmentation, 3-D reconstruction and projection. It considers the differences in surface reflection in different image regions, separates the image into Lambert surface and specula surface, reconstructs the shapes respectively, then fuses the two results, changes the illuminant direction and power, generates a new image. The second method fuses the shape from shading and shape from texture techniques. Gabor filters are used to separate the shading information and texture information, then 3-D images are reconstructed respectively, the two 3-D images are fused according to their characters in frequency field and the new 2-D image is generated by illumination changing. The experiments show that the images generated by those two methods are true, and better than that generated by traditional methods.