调整量最小的多视图像校正算法

The Adjustment Minimizing Multi-view Image Correction Algorithm

  • 摘要: 在多视自由视点电视(FTV)系统中,摄像机阵列拍摄的多视视频存在垂直视差、且水平视差分布不均等问题,严重影响合成视频的质量效果。现有的多视图像校正方法虽然消除了垂直视差,但是校正图像过程中引入的旋转和平移失真较大,使人眼产生不舒适感。为此,本文提出一种适于平行摄像机阵列的多视图像校正算法,首先建立理想的摄像机安放位置,然后计算多视图像校正变换,将实拍图像映射到理想摄像机位置处,得到校正后图像。在设计理想的摄像机安放位置时,考虑人眼视觉特性,确保坐标轴旋转和平移调整量最小化,实现多视图像校正变换引入图像失真总量最小。实验表明,该算法不仅可以消除垂直视差,而且图像校正过程中引入的失真小,符合人眼视觉习惯,能有效改善合成视的图像质量,提高人眼观看舒适程度。

     

    Abstract: These manually camera set may cause vertical parallax and uneven horizontal parallax in multi-view images, which seriously affects the quality of synthesized video. A multi-view image correction algorithm for parallel camera array is presented in this paper. According to HVS characteristics, an ideal multi-camera set is designed based on the constraint of minimum adjustment. The experiments show that the algorithm can reduce the vertical parallax greatly with introducing less distortion in the corrected multi-view images. It is consistent with human visual habits and effectively improves the image quality of the synthesized video.

     

/

返回文章
返回