Thumb Recognition System using Polar Harmonic Transform(PHT) for Rotation Invariance

Project Description

We develop a fast approach for their computation using recursion and 8-way symmetry/anti symmetry property of the kernel functions Polar harmonic transforms (PHTs) are orthogonal rotation invariant transforms that provide many numerically stable features. The kernel functions of PHTs consist of sinusoidal functions that are inherently computation Polar harmonic transform (PHT) which can be used to generate rotation invariant features. With PHTs, there is also no numerical instability issue, as with ZM and PZMs which often limits their practical usefulness. A large part of the computation of the PHT kernels can be recomputed and stored. In the end, for each pixel, as little as three multiplications, one addition operation, and one cosine and/or sine evaluation are needed to obtain the final kernel value. In this project, three different transforms will be introduced, namely, Polar Complex Exponential Transform (PCET), Polar Cosine Transform (PCT), and Polar Sine Transform (PST).