應用粒子群演算法於區域導向影像定位效能評估

Performance Evaluation on Area-Based Image Registration Using Particle Swarm Optimization

連昇鴻(Sheng-Hung Lien)

指導教授 ： 范書楷

國立臺北科技大學/管理學院/工業工程與管理系碩士班/碩士(2012年)

https://doi.org/10.6841/NTUT.2012.00486

摘要

影像定位主要是透過兩張或多張影像間共同的資訊將影像結合，這些影像可以由不同的觀測點、不同的時間或是不同的感應器所取得。影像定位主要可以分為四個步驟：影像特徵偵測、特徵配對、轉換方程式估計轉換量以及影像修補。而影像定位之方法主要可分為特徵導向(feature based)和區域導向(area based)。本論文所討論的影像定位方式是區域導向，透過交叉相關性(Cross-correlation)、共同資訊量(Mutual Information)找到影像間相似性最高的區域。本研究中將利用粒子群演算法對轉換方程式估計轉換量進行最佳化，把粒子群演算法之參數進行參數實驗設計，進而提供參數給粒子群演算法應用於影像定位，以提高影像定位之準確率，在此將測試五種類別影像分別為：醫療影像、自然影像、空照圖、大位移影像、高對比影像；尋找一組參數組合均適用於交叉相關性(Cross-correlation)、共同資訊量(Mutual Information)、標準化共同資訊量(Normalize Mutual Information) 三種目標函式方法，最後再與先前應用粒子群演算法於影像定位的文獻比較成功率以及運算效率。

關鍵字

影像定位 ； 粒子群演算法 ； 交叉相關性 ； 共同資訊量 ； 實驗設計

並列摘要

Image registration is an important image processing step of aligning two or more images by means of their common information. These images should be taken from different viewpoints, times, or sensors. The procedure of image registration consists of four basic steps: feature extraction, feature matching, transformation model estimation, and image resampling. Image registration methods are classified as feature based and area based methods. The area based methods will be the focus in this thesis. Most cross-correlation and mutual information methods are developed based on image intensities for the direct matching purpose. In this thesis, particle swarm optimization will be utilized to solve the transformation model, and the design of experiments will be conducted for particle swarm optimization, providing a parameter set to improve the registration rate. All the methods considered will be tested for five categories of image sets: medical image, nature image, remote sensing, large translation image and high contrast image. The parameter setting well adapted to cross-correlation, mutual information, and normalize mutual information objective function methods is investigated. At last, the comparison results among the studied methods are reported in terms of the success rate and the number of function evaluations.

並列關鍵字

Image Registration ； Particle Swarm Optimization ； Cross-correlation, Mutual Information ； Normalize Mutual Information ； Design of Experiments

參考文獻

J. Canny, “A computational approach to edge detection, ” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 8, no.1, 1986, pp. 679-698.

A. Goshtasby, “Description and discrimination of planar shapes using shape matrices,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.7, no.1, 1985, pp. 738-743.

L. Kitchen, A. Rosenfeld, “Gray-level corner detection,” Pattern Recognition Letters, Vol. 1, no. 1, 1982, pp. 95-102.

E. G. Keys, (1981). “Cubic convolution interpolation for digital image processing,” IEEE Transactions on Acoustics, Speech, and Signal Processing, Vol. 29, no. 6, 1981, pp. 1153-1160.

T. M. Lehmann, C. Gonner, Spitzer, K. (1999). “Survey: interpolation methods in medical image processing,” IEEE Transactions on Medical Image, Vol.18, no.1, 1999, pp. 1049-1075.