Linde-Buzo-Gray (LBG) algorithm is used in image processing for Vector Quantization (VQ). LBG
technique is robust, performs locally best but depends on the initial codebook. In the splitting based VQ, the first
center is defined as average of all vectors. The rest of 2n centres are calculated by splitting and updated procedure.
In the proposed new technique (NLBG) the LBG is improved and insted of splitting all centres into two new
areas, the worst area that has highest mean square error splitted and updated into to new areas. Therefore, the
number of codevectors is increased one by one apart from the classical LBG. Consequently, the performance of
the codebook is increased globally. In this paper, the new technique is applied to the standard images, compared
to the FCM(Fuzzy C-Means), K-Means (K – Ortalamalar) and LBG. As a result, it is seen that the proposed new
technique performs better according to the criteria of MSE.
Birincil Dil | Türkçe |
---|---|
Konular | Mühendislik |
Bölüm | Elektrik Elektronik Mühendisliği / Electrical Electronic Engineering |
Yazarlar | |
Yayımlanma Tarihi | 31 Mart 2017 |
Gönderilme Tarihi | 16 Mayıs 2016 |
Kabul Tarihi | 23 Ağustos 2016 |
Yayımlandığı Sayı | Yıl 2017 Cilt: 7 Sayı: 1 |