CAJ | 학술논문

为了实现对多视点图像的融合，提出了一种使用分块广义PCA（ GPCA）的方法。分块可以将图像处理的过程细化，简化计算GPCA则考虑了二维数据的空间关联性，用于灰度图像降维时有较好的效果，两者的结合是文章的一个创新。由于需要考虑常规多视点图像的不同视点间存在位移差的事实，图像的预处理环节加入了必要的配准和投影变换操作。因此，整个方法主要包括图像匹配、投影变换、分块GPCA计算和融合等环节。为验证方法的可行性和准确性，文章引入了二维经验模态分解BEMD。结果表明，和BEMD相比，所提方法在图像融合的性能和计算复杂度上都表现出了优势，有一定的实用价值。
위료실현대다시점도상적융합，제출료일충사용분괴엄의PCA（ GPCA）적방법。분괴가이장도상처리적과정세화，간화계산GPCA칙고필료이유수거적공간관련성，용우회도도상강유시유교호적효과，량자적결합시문장적일개창신。유우수요고필상규다시점도상적불동시점간존재위이차적사실，도상적예처리배절가입료필요적배준화투영변환조작。인차，정개방법주요포괄도상필배、투영변환、분괴GPCA계산화융합등배절。위험증방법적가행성화준학성，문장인입료이유경험모태분해BEMD。결과표명，화BEMD상비，소제방법재도상융합적성능화계산복잡도상도표현출료우세，유일정적실용개치。
In order to complete the fusion of multiview images, this paper presents a generalized block-based PCA ( GPCA) method. Blocking can refine the progress of image processing and simplify the calculation. GPCA consid-ers the spatial correlation between two-dimensional data and has better results in dimensinality reduction of gray im-ages, so it is an innovation by combining these two methods in this article. Because there has to be recognition of the fact that there exists differential displacement between the different viewpoints of the conventional multi-view im-age, some necessary registration and projection transformation operations are added to the pre-processing of the ima-ges. Thus, the entire method includes image matching, projection transformation, computing of GPCA, image fu-sion and so on. In order to verify this method's feasibility, this article has introduced bi-dimensional empirical mode decomposition ( BEMD) . The results show that compared with BEMD, GPCA demonstrates certain advantages in performance and computational complexity regarding image fusion, and has some practical value.