动目标多视角观测图像的三维去模糊方法

当目标在多视角观测成像系统中有相对运动时,所获取的多观测点图像是模糊的且各视角图像的模糊是不一样的,模糊核长度和方向都不同。针对这一问题,提出了多视角观测图像的三维去模糊方法。现有的图像去模糊方法主要是对单视角观测图像进行二维去模糊的,没有考虑目标多视角观测图像的模糊核之间的对应关系。文中从三维空间到二维观测面的映射关系出发,建立多视角观测图像的模糊路径之间的对应关系。先采用单观测图像去模糊的方法获取两视角观测图像的模糊核,并对模糊核进行精确化处理,得到单像素点宽的模糊核路径。再通过多视角观测图像模糊核路径之间的对应关系,估计其它观测图像的模糊核路径。最后,对多视角观测图像进行统一去模糊,并对去模糊后的多视角观测图像进行三维重建。实验结果表明,文中方法能较好地去除目标多视角观测图像的三维模糊,提高了目标的三维重建质量。

Three-dimension deblurring algorithm for multiple observed images of moving object

For the problem of deblurring of multiple observed images of moving objects with different blur kernels, a joint three-dimension deblurring method for multiple observed images was proposed. Unlike existing deblurring methods to remove 2D blur for single observed image without considering relationship of blur kernels in multiple observed images, the relationship of different PSF paths of multiple observed images was explored. The movement in three-dimension space was projected into each observed image planes, hence, the inherent relationship between the two PSF paths of two observed images can be set up. At first, motion blur kernels of two view images were estimated using single observed image debluring method and the blur kernels with one pixel width was refined. When having estimated two PSF paths, the other PSF paths could be calculated and optimized by using the relationship of PSF paths in multiple observed images. Then 3D blur of the observed images could be removed by using the proposed method. The experiment results for multiple observed images demonstrate that the proposed approach is efficient and effective in removing 3D blur and reconstruction.