光场图片中基于Plenoptic的尺度不变特征描述符

Plenoptic based scale-invariant feature descriptor in light field images

为了使光场图片在旋转和缩放等操作下保持不变,基于Plenoptic函数的图片表达方法提出一种尺度不变特征描述符。首先,对Plenoptic函数与光场进行了描述,并用七维数据来表示时序图片序列;其次,对Plenoptic函数的规范化进行了分析,并给出了三维场景点向二维和三维Plenoptic空间的映射方法;接下来,分析了光场中图片在旋转和缩放操作下的尺度不变特性,并给出了相应的变换算法;最后,在提取尺度不变的特征向量时通过融合函数降低了参数调整所需的计算量。实验表明,提出的尺度不变特征描述符可以更好地描述图片,从而在基于描述符的操作中具有更优的性能。

In order to keep scale-invariant of an image in light field under rotation and scaling,this paper proposed a scale-invariant feature descriptor for images based on Plenoptic function. Firstly,it analyzed Plenoptic function and light field,and used a seven-dimensional data to represent time-ordered sequence of images. Secondly,it analyzed the regularity of the Plenoptic function,and gave that how to map a 3-dimensional point in a scene to a 2-dimensional or 3-dimensional plane in Plenoptic space.Thirdly,it analyzed the scale-invariant features of images under rotation and scaling in light field,and proposed corresponding transform algorithms. Finally,while extracting scale-invariant vector,it decreased the computation with proposed integration function. The experiments show that,the proposed scale-invariant feature descriptor can better represent an image,and has better performance than related works.