Published June 30, 2021 | Version v1

COMPARISON OF WAVELET TRANSFORM FOR IMAGE RECOGNITION SYSTEM USING LEARNING VECTOR QUANTIZATION

Authors/Creators

  • 1. Jurusan Teknik Informatika, Universitas Halu Oleo Kendari

Description

Image is a spatial dimension contains information, color, and not time-dependent. Nowadays image is very important for recognition system as source/data. In order to obtain certain information (features), image transformed or extracted.  Wavelet is mathematical function that is able to classify the image energy concentrated on a small group of coefficients (approximation), while other coefficient group (detail) contains very small energy which can be eliminated. This study purpose to compare wavelet family include Haar, Daubechies and Coiflet, then applied to the fingerprint recognition system and face recognition system.

There are three major steps in this paper, preprocessing that includes resize normalization, histogram equalization and thresholding, feature extraction using three different wavelet family, learning using Learning Vector Quantization (LVQ) and matching using Euclidean Distance. The experimental result show Haar is the best among the others. In Fingerprint recognition system the accuracy is 85% and 90% for face recognition system

Files

1_vol 7 no 1 hal 1-8 Rizal Adi Saputra.pdf

Files (602.2 kB)

Name Size
md5:0df720cfa2d715526357d247af3ab687
602.2 kB Preview Download