Image reconstruction: A comparison between moment and non-moment based techniques

This paper presents a comparison between moment and non-moment based techniques in image reconstruction. The moment based technique used is Zernike Moments (ZMs) while Fast Fourier Transform (FFT) is the non-moment based technique. Considering the inverse process of these two techniques will allow the reconstruction of the image. Two types of images are considered namely gray scales and binary and the original images are corrupted with three different noises that are Salt and Pepper, Gaussian and Random. The performance of each algorithm against original and corrupted images is measured by evaluating Peak Signal to Noise Ratio (PSNR) for each reconstructed gray scale image of pixel size 64×64 and binary image of pixel size 30×30. The PSNR for each type of images will be observed in terms of the effectiveness of both Inverse Fast Fourier Transform (IFFT) and Inverse Zernike Moment (IZM) as reconstruction algorithms. The results show that between the two techniques, ZM is less sensitive to noise compared to FFT even though FFT is able to reconstruct a reasonably quality image.

I. Introduction

Image reconstruction is very much concerned with the ability to reproduce image of possibly the same quality as original. Image reconstruction basically is useful in many areas for instance reconstruction of medical images, in Biometrics such as fingerprint reconstruction for security purposes, face image reconstruction for recognition and many more. Unlike image enhancement, image reconstruction is the attempt to retrieve information that has been lost or obscured in the image processing itself [1].