An Empirical Analysis of Invariance Hu's Moment Feature over a Digital Image

The invariant moment feature of Digital images provides a robust feature in the domain of computer vision to pattern classification and recognition for the last many years. Hu's seven invariant moments have a unique capability of image recognition under image transformation like translation, scaling, and rotation. Originally, Hu's moments were proposed under the assumption that the images are continuous domain and noiseless, but in the case of digital image moment, invariant features no longer remain perfectly invariant. This paper proposed a detailed analysis of these moments on digital images and observed that the invariance property is slightly degraded due to discretization in digitized images.