基於疊代法多次分割圖像應用在閥值分割

Image segmentation by Iterative Method with multi-split

許育棋(Yu-Chi Hsu)

指導教授 ： 林啟芳

元智大學/資訊學院/資訊工程學系/碩士(2009年)

摘要

閥值分割(thresholding)是影像處理中常見的前處理(preprocessing)步驟，分割結果的好壞往往影響後續處理的準確度。常用的分割演算法有最大類間方差(variance of between class)法、閥值分割疊代(iterative)法、最大熵值(entropy)法、中心群聚(cluster)分割法、模糊(Fuzzy)閥值分割法等，而處理的區域範圍又可分為全區域閥值分割與局部區域閥值分割。一般而言，上述演算法比較適合應用在大面積目標物的提取(extraction)，但如果所要處理的目標物較小，例如在MRI (magnetic resonance imaging)影像裡的顯影劑目標，或在衛星影像中的細長道路，其目標物佔整體影像的面積比例較小，所得分割效果往往很差。原因是這類目標物的像素太少，容易被錯分為背景。但這類目標物大多具有高亮度的特徵，因此本論文提出一個全區域多次分割閥值法，以簡單步驟來提取這類小面積且具高亮度特徵的目標物，提供給後續步驟作進一步處理。

關鍵字

直方圖 ； 疊代法 ； 閥值

並列摘要

Gray level thresholding it is image prepositive process,a commonly used one perform algorithm have Otsu ,Iterative Method, Tsallis entropy. And the histogram regional range with global regional and dynamic regional, generally speaking can get the good result on above-mentioned performing algorithms and applying the treatment that is gray level thresholding to the image.
But this kind of method drew the smaller goal thing, for instance: Small goal developer, road to the satellite image in MRI image of medical treatment, the results of cutting apart of its background and goal thing are bad, its reason is looking like prime and few for the characteristic of this kind of goal thing, it is apt to make the goal thing divided into the background by mistake.
Developer part (generally small blocks) of the picture MIR image , road, satellite of image take image to be area getting little, it is difficult for the goal thing to cut apart, and this kind of goal thing is mostly high light (for image) Attribute, utilize goal to be thing little on the basis of the thesis having luminance high characteristic, propose drawing the goal thing by way of cutting apart many times, after abstraction of valve value, characteristics including goal thing are the more, and irrelevant background is the fewer in two value images of the income, have healed and can reduce the wrong result for the follow-up treatment procedure step.

並列關鍵字

Histogram partition ； Iterative method ； Thresholding

參考文獻

[1] N. Otsu, “A threshold selection method from gray level histograms,” IEEE Transactions on Systems, Man, and Cybernetics, pp. 62–66, 1979.

[2] T. W. Ridler and S. Calvard, “Picture thresholding using an iterative selection method,” IEEE Transactions on System, Man, and Cybernetics, pp. 630–632, 1978.

[4] M. Portes, I. A. Esquef, A. R. Gesualdi Mello, and M. Portes, “Image thresholding using Tsallis entropy, ” Pattern Recognition Letters, 25, pp. 1059-1065, 2004.

[5] J. N. Kapur, P. K. Sahoo and A.K.C. Wong, “A new method for gray-level picture thresholding using the entropy of the histogram,” Computer Vision, and Graphics Image Processing, Vol. 29, pp. 273–285, 1985.

[7] C. Chinrungrueng and C. H. Sequin, “Optimal adaptive K-means algorithm with dynamic adjustment of learning rate,” IEEE Transactions on Neural Networks, Vol. 6, N0. 1, pp. 157–169, 1995.