Phase-only Wavelet Matched Filtering for Two- Dimensional Image Recognition and Classification

Abstract

Objectives: The adaptability of the Wavelet Matched Filtering which uses wavelets along with Fourier transforms is applied for the recognition and classification of two-dimensional objects. In image processing and pattern recognition, phase information is very relevant as it contains the features of the visual scene. Methods/Statistical Analysis: The paper describes the design for the phase-only wavelet matched filtering. The performance of the phase-only wavelet matched filter is compared with that of wavelet matched filter, the classical Matched Filter and the Phase-only Matched Filter for classifying four objects in a 2D scene. The comparison is done using the MATLAB simulation tool. Findings: Two metrics - the Peak to Correlation Energy (PCE) and the Discrimination Ratio (DR), were used to study the performance of Phase-only Wavelet Matched Filtering (PWMF), Wavelet Matched Filtering (WMF), Phase-only Matched Filtering (PMF) and classical Matched Filtering (MF). It is found that the Phase-only Wavelet Matched Filtering (PWMF) gives better discrimination capability and the proof of the concept is demonstrated here. Applications: The method finds its application in pattern or object recognition and classification.