Name
Abstract | ||
|---|---|---|
This paper presents a novel method to derive invariants of 2D grayscale images under projective transformation. Invariants of images are good features for object recognition and have attracted extensive attention. Although geometric invariants of point locations such as cross ratios are well known for centuries, we have found no reported invariants for grayscale images that remain the same under projective transformation. It has even been proven that projective invariants of images cannot be derived from the standard geometric moments of images. However, this does not mean that there is no projective invariant of images in other forms. We will prove in this paper that projective invariants of images do exist as functions of the generalized moments of images. We first derive some projective invariant relations between an image function and its derivative functions. Next, we extend the traditional definition of moments by considering both the image function and its derivative functions. Then we derive a set of functions of the generalized moments that are projective invariant. Experimental results indicate that the proposed invariants have certain discriminating power for object recognition. |
Year | DOI | Venue |
|---|---|---|
2015 | 10.1007/s10851-014-0518-z | Journal of Mathematical Imaging and Vision |
Keywords | Field | DocType |
Derivative image,Invariant,Moment,Projective transformation | Twisted cubic,Projective geometry,Homography,Artificial intelligence,Collineation,Computer vision,Discrete mathematics,Projective differential geometry,Projective harmonic conjugate,Pure mathematics,Pencil (mathematics),Mathematics,Projective space | Journal |
Volume | Issue | ISSN |
51 | 2 | 0924-9907 |
Citations | PageRank | References |
0 | 0.34 | 41 |
Authors | ||
4 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Yuanbin Wang | 1 | 22 | 3.76 |
| Xingwei Wang | 2 | 1025 | 154.16 |
| Bin Zhang | 3 | 213 | 41.40 |
| Ying Wang | 4 | 0 | 0.34 |