應用混種式最佳化演算法於廣義高斯混合模型參數估計之研究

Parameter Estimation of Generalized Gaussian Mixture Model Using a Hybrid Heuristic Method

吳佳蒨(Chia-Chan Wu)

指導教授 ： 范書愷

元智大學/工程學院/工業工程與管理學系/碩士(2006年)

https://doi.org/10.6838/YZU.2006.00226

摘要

廣義高斯混合（generalized Gaussian mixture）模型是一種常被使用來呈現訊號處理其統計特性的參數統計模型。此模型可以良好的配適自然圖像的轉換係數，因此在此論文中我們採用其去近似原始資料的組合分配。並且提出一個混種式最佳化演算法(hybrid heuristic algorithm)對此廣義高斯混合模型做配適和對其參數做估計，尤其是針對此模型較大值的形狀參數(shape parameter)。此混種式最佳化演算法是結合質群演算法(Particle Swarm Optimization)和Entropy Matching Estimator(EME) 方法去尋找此組合分配的最佳參數估計值。在實驗研究上，我們將此演算法和假設模型應用在影像分割處理(image thresholding)上。影像分割處理是一種廣泛被使用來區分主體和背景的方法。我們使用假設模型和演算法去近似影像的機率分布圖並求出其影像的整體臨界值(global thresholding)。從實驗結果中可以呈現出此結合式演算法在解決多階的影像臨界值問題上有較好的效果並且能成功地完整配適非高斯(no-Gaussian)的影像機率分布圖。

關鍵字

廣義高斯混合模型；質群演算法；Entropy Matching Estimator (EME)方法；影像分割處理；形狀參數。

並列摘要

The generalized Gaussian distribution mixture (mixed GGD) model is a parametric statistical model, which is often used to characterize the statistical behavior of a process signal. This thesis adopts the mixture GGD model to approximate the associated empirical distributions, which could be best fitted to the discrete cosine transform (DCT) coefficients of all natural images. A heuristic method is developed to estimate the parameters of the mixture GGD. The proposed heuristic method is designed for fitting the mixture GGD, particularly for the large shape parameter in the mixture GGD. The heuristic method integrates the Particle Swarm Optimization (PSO) algorithm and the Entropy Matching Estimator (EME) method to seek the optimal estimates of the distribution. The experimental study in terms of image segmentation is presented to illustrate the proposed method. Image thresholding is a useful technique to separate out the interested object from background information. So, we adopt the parametric approach method by fitting the intensity of image to find the Optimal thresholding. It is assumed that the intensity of image can be well represented by the mixture GGD model. The experimental results show that the integrated heuristic method could provide better effectiveness in the case of the multi-level thresholding problem and depicts quite successfully the non-Gaussian probability density function (PDF) of image intensity.

並列關鍵字

generalized Gaussian distribution mixture (mixture GGD) model ； Particle Swarm Optimization (PSO) algorithm ； Entropy Matching Estimator (EME) ； Images thresholding ； Shape parameter.

參考文獻

Aiazzi, B., Alparone, L. and Baronti, S. “Estimation Based on Entropy Matching for Generalized Gaussian PDF Modeling”, IEEE Signal Processing Letters, Vol. 6, No. 6, pp. 138-140, June 1999.

Aiazzi, B., Baronti, S., Santurri, L., Selva, M. and Alparone, L. “Information-theoretic Assessment of multi-dimensional Signals”, Signal Processing, Vol. 85 pp. 903-916, 2005.

Birney, K. A. and Fisher, T. R. “On The Modeling of DCT and Subband Image Data for Compression”, IEEE Trans on Image Processing, Vol. 4, No. 2, pp. 186-193, February 1995.

Brits, R. “Niching Strategies for Particle Swarm Optimization”, pp. 13-16, November 2002.

Bazi, Y., Bruzzone, L. and Melgani, F. “An Unsupervised Approach Based on The Generalized Gaussian Model to Automatic Change Detection in Multitemporal SAR Images”, IEEE Transactions on Geoscience and Remote Sensing, Vol. 43, No. 4, April 2005.