Abstract:
Object recognition is one of the biggest problems in computer vision. In order to be able to solve this problem, many methods have been tried in the past. Wavelets, which have successfully been used in such areas as image compression, and denoising, have been applied to the object recognition problem as well. When planar objects are photographed from different view points, they seem to be going through an affine transformation. Therefore, an affine invariant wavelet function has been formulated in a recent paper for recognition of planar objects. This work aims to implement the proposed invariant function with the multiwavelet transform to improve its results, for multiwavelets are known to have superior performance compared to scalar wavelets in other signal processing fields. A comparison of the changes in the performance of the affine invariant function when it is implemented with scalar and multiwavelets is done. The function has also been implemented with the combined waveletcoefficients obtained by using two different set of filters. A classical method, Fourier descriptors have been included in the simulations for comparison purposes as well.
