Polar Representation of 2D Image Using Complex Exponential Spiking Neuron Network

Abstract

The paper introduces an innovative hybrid encoding method for images. It proposes a conversion process where the image is transformed from the conventional Cartesian coordinates representation (x and y) to a polar coordinates representation using complex numbers with magnitude ρ and angle θ. By doing so, the spatial information of pixel locations and the connections among adjacent pixels are preserved during the serialization process, and the frequency components of the image are effectively calculated and combined during Discrete Fourier Transform. Furthermore, the paper outlines the utilization of a spiking neural network (SNN) for representing and reconstructing the image. The SNN is constructed by combining the outputs of 128 complex exponential spiking neurons. Each spiking neuron corresponds to a specific frequency component of the image. By employing this hybrid encoding method and utilizing the SNN, the research aims to enhance image representation and reconstruction processes.