ABSTRAKSI: Permasalahan wavelet thresholding pada image denoising adalah
bagaimana menentukan nilai threshold yang tepat. Penggunaan metoda seperti
Normalshrink untuk mencari nilai threshold bisa menyelesaikan permasalahan.
Namun metoda Normalshrink mengasumsikan bahwa wavelet coeffient bersifat
independent. Penggunaan metoda Bivariate Shrinkage Dengan Local Variance
Estimation tetap mempertahankan sifat dependent dari wavelet coefficient
sehingga bisa meningkatkan performansi image denoising. Performansi metoda
ini dipengaruhi oleh ukuran windowsize pada saat perhitungan variance dari citra
ternoise.
Dalam Tugas Akhir ini telah dianalisis dan diimplementasikan image
denoising menggunakan metode Bivariate Shrinkage Dengan Local Variance
Estimation. Pengujian dilakukan terhadap berbagai ukuran windowsize sehingga
diketahui pengaruhnya terhadap PSNR hasil denoising dan waktu komputasi
proses denoising. Noise yang digunakan dalam pengujian adalah additive
gaussian noise, additive laplacian noise, dan impulsive noise yang dibangkitkan
melalui suatu noise generator.
Dari hasil percobaan didapatkan bahwa metoda Bivariate Shrinkage
Dengan Local Variance Estimation mendapatkan PSNR hasil denoising yang
lebih baik sekitar 0.01~0.5 dB terhadap Bivariate Shrinkage dan 0.05~1.5 dB
terhadap Normalshrink. Waktu komputasi proses denoising metoda ini
dipengaruhi oleh ukuran windowsize, semakin besar windowsize maka semakin
tinggi waktu komputasi proses denoising.Kata Kunci : wavelet thresholding, image denoising, bivariate shrinkage, local variance estimation, windowsize.ABSTRACT: The main problem in image denoising using wavelet thresholding is how
to obtain the effective threshold value. The Normalshrink usage to obtain this
value can be accomplish the problem. But Normalshrink assumes that wavelet
coefficients are independent each other. Bivariate Shrinkage With Local Variance
Estimation usage keeps the dependent between wavelet coefficient so can improve
the performance of image denoising. The performance of this method is
influenced by windowsize in noised image’s marginal variance measurement
In this Final Project, it has been analysed and implemented the used of
Bivariate Shrinkage With Local Variance Estimation method for image denoising.
Testing phase is toward to varying windowsize so the influences in denoising
PSNR’s result and computational time will be known. The noise which is used in
testing phase are additive gaussian noise, additive laplacian noise and impulsive
noise which is generated by noise generator.
From the experiment result, Bivariate Shrinkage With Local Variance
Estimation method have better PSNR’s denoising result about 0.01~0.5 dB toward
to Bivariate Shrinkage and 0.05~1.5 dB toward to Normalshrink. Denoising
computational time of this method is influenced by windowsize, bigger
windowsize needs bigger denoising computational time.Keyword: wavelet thresholding, image denoising, bivariate shrinkage, local variance estimation, windowsize.