Moment-Based Image Analysis And Fast Computation Of Moment Invariants

Moment function is an important tool in image analysis and has been widely used in the field of computer vision, image processing and pattern recognition. Examples of moment based feature descriptors include Cartesian geometrical moment, orthogonal moment and wavelet moment. Cartesian geometrical moment, because of its simple form, has been well studied. In recent years, moment with an orthogonal basis set (e. g., Legendre and Zernike polynomials) becomes more active because of their good properties. They have small noisy sensibility, can be applied to reconstruct object directly and represent image with a minimal amount of information redundancy. But these kinds of moments can only extract global features from images, so wavelet moment appeared. The most important advantage of it is that wavelet moment can extract local features from images, so it is a very useful descriptor in pattern recognition.This dissertation mainly focuses on moment based image analysis and application, including some algorithms for efficient computation of Legendre moments, moment-based method for template matching, and the application of wavelet moment in pattern recognition.On the aspect of fast algorithm, we use image blocks to represent images, and then propose two new algorithms for fast computation of Legendre moment, which are integral method and cumulative method respectively.As to template matching, we use a two stage method. In the first stage, the matching candidates are selected using a computationally low cost feature. Frequency domain calculation is adopted to reduce the computational cost for this stage. In the second stage, rotation invariant template matching is performed only on the matching candidates using Zernike moments. This algorithm is very fast and accurate.We also did some work on the application of wavelet moments. We choose the-II-cubic B-spline wavelets and Haar wavelets to construct wavelet moments respectively. They are not only invariant to translation, scaling and rotation but also have the multi-resolution properties which are suitable for classifying very similar objects. Because of their advantages, we use them in Chinese characters recognition. The results show that wavelet moment is really very useful and better than Zernike moment.