快速分割的改良技術和邊界資訊的壓縮

Improvement Techniques for Fast Segmentation and Compression for Boundary Information

洪文琦(Wen-Chih Hong)

指導教授 ： 丁建均

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.00411

摘要

此論文之當然目標為發展一種圖像之切割技術，其結合快速掃描演算法、圖像型態學運算子抑或若干幾何數學概念；同時提出一種新穎的邊界之描述技術以便能將切割後之區塊邊界作記錄而後壓縮。
圖像切割方面，我們研究並針對快速掃描切割技術提出了兩項截然不同的改良技術。第一項改良法利用了圖像型態學的特性：侵蝕、擴張。灰階圖像的侵蝕及擴張存在著使圖像模糊化及邊界平滑化之能力。如此可幫助圖像切割法針對圖像能有更加理想的切割結果。且二位元侵蝕及擴張也可使邊界平滑化，還可消除雜訊，最重要的是可以將一些不該歸屬成同區的區塊分成兩塊以上。
另一項改良法則是用一多邊形模擬一個區塊，並關注此一多邊形的頂點內角。理論上，頂點內角的角度如果大於一百八十度，則鄰近此頂點的邊界為凹邊界；相反地，頂點內角角度大小於一百八十度，則鄰近此頂點的邊界為凸邊界。而如果一個多邊形存在多個凹邊界，我們則測量位於不同凹邊界之頂點之間的最短距離。此一測量用於決定是否連接此最短距離的兩凹邊界頂點。
再次強調，邊界描述技術是用來記錄圖像切割後之各區塊的邊界輪廓。能用越少量的位元組描述邊界並越完美重建邊界輪廓，可稱為即佳的邊界描述技術。
我們提出的邊界描述技術為運用多項式方程來近似邊界。首先，必須先將封閉的邊界分割成若干開放的小片段。接著利用二階的多項方程式來fit 。最後再將這些二階多項式方程依序頭尾連接成一個封閉的邊界。
無論切割技術或邊界描述技術，最後都有一連串的實際模擬結果。

關鍵字

影像切割 ； 邊界描述

並列摘要

The goal of this thesis is to develop image segmentation methods which combine fast scanning algorithm with morphological operations and some geometrical processing. Meanwhile, a new boundary descriptor method is proposed to record and compress the boundary data of region among segmentation results.
In image segmentation, there are two improvements be proposed. One is based on the morphological characteristic of erosion and dilation. Grayscale erosion and dilation could blur and smooth an image. It helps segmentation method more easy to classify some interruptions into one region, for example, a high variation part. While binary erosion and dilation could smooth boundary, remove noise, and divide some region into two regions.
Another improvement is based on the characteristic of inner angles of polygons. Degree of inner angles of a polygon usually corresponds to how it corresponding curve is concave or convex. If many concave parts existed in a closed boundary, we can measure and check the distances of different concave parts in order to decide whether a connection needed or not.
A boundary description is used to record the boundaries of every regions of segmented image. Of course, an excellent description is using less bytes to record data from which boundary could be reconstructed.
A new technique with concept of polynomial approximation is developed. It divides a closed boundary into many sections, then using a second order polynomial function to fit corresponding section. In the end, rearranging the second order polynomial functions.
No matter segmentation or boundary description, the experimental results are given.

並列關鍵字

Image segmentation ； boundary description

參考文獻

[2] R. Adams and L. Bischof, "Seeded region growing," in IEEE Trans. Pattern Anal. Mach. Intelligence, vol. 16 no.6, pp. 641-647, June 1994.

[4] R. M. Haralick and L. G. Shapiro, "Image segmentation techniques," in Comput. Vision Graphics Image Process., vol. 29, pp.100-132, 1985.

[6] Dorin Comaniciu and Peter Meer, " Mean Shift: A Robust Approach Toward Feature Space Analysis, " in IEEE Trans. Pattern Analysis and Machine Intelligence, vol.24, no.5, pp.603-619, May 2002.

[7] Yizong Cheng, "Mean Shift, Mode Seeking, and Clustering," in IEEE Trans. Pattern Analysis and Machine Intelligence, vol.17, no.8, pp.790-799, August 1995.

[9] M. V. Sudhamani and Dr. C. R. Venugopal "Segmentation of Color Images using Mean Shift Algorithm for Feature Extraction, " in IEEE 9th International Conference on Information Technology, pp.241-242, 2006.