On digital approximation of moment invariants
Computer Vision, Graphics, and Image Processing
Generation of moment invariants and their uses for character recognition
Pattern Recognition Letters
Moment Invariants and Quantization Effects
CVPR '98 Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
| Hi-index | 0.01 |
For multimedia retrieval application, shape is always a conspicuous element of an object. Moment-based approaches are widely used for shape description due to its translation, scaling and rotation invariance. Moment invariants are defined in the continuous domain. However, when considering the digital images in practice, quantization errors are introduced. Thus, the moment invariants calculated might not be truly invariant. This paper presents an analysis of quantization effects on four moment-based approaches of both regular and irregular objects. From the analysis, the scaling errors for all approaches are large when the scaling factor is less than 0.5. Moreover, the rotational errors are big for the objects rotated other than the multiples of 90掳. Our experimental results show that Dudani moment invariants suffer the largest error for overall sensitivity, while Affine moment invariants show the smallest. Furthermore, this error analysis has also been applied successfully to object searching applications using a threshold selection scheme.