基於分水嶺演算法之影像區域表示法

The Image Region Representation Based on Watershed Algorithm

蔡榮峯(Rong-Fong Cai)

指導教授 ： 黃健興

義守大學/電機資訊學院/資訊工程學系/碩士(2014年)

https://doi.org/10.6343/ISU.2014.00176

摘要

隨著科技的進步，數位影像資料的使用越來越頻繁，依據不同的應用衍生許多種不同的影像表示方法，在不同的影像應用中，搭配的影像表示方法會影響其效能。在此論文中，提出一種以影像內容為基礎的影像表示方法，以影像色彩構成區域為單位，保留色彩紋理資訊，最後以內插法重建影像色彩內容。
我們採用分水嶺演算法，以色彩變化梯度方向為紋理資訊，對影像色彩分佈進行分割；接著使用非對稱與反包裝模式表示影像內容，以梯度方向表示色層變化趨勢，則影像可分解為多個色調區域。記錄每個區域中影像強度極值與梯度變化方向，其餘位置的影像強度可透過線性內插法計算。
經實驗驗證，限制梯度變化量會影響分水嶺切割與色彩重建結果，當影像強度變化趨勢平緩，可使用較少的色彩構成區域來表示影像，且色彩重建時失真度較小，反之則需要更多的色彩構成區域來表示影像內容。

關鍵字

影像分割 ； 影像表示 ； 影像重建 ； 分水嶺演算法 ； 非對稱影像表示法 ； 斜面分解表示法 ； 分層表示法

並列摘要

Advancement of technology is most frequently used digital image data . There are many different image representation developed for several kinds of image application. The suitable represent method would improve the image applications. In this thesis, we propose a method to represent the image based on the image color constitute region, to retain the color texture information, and to reconstruct image by interpolation.
First, we segment the image by the watershed algorithm to get color gradient direction as the texture information, and then represent the image content by asymmetric and anti-packing model. The gradient direction would indicate the color trend layer and the image would be decomposed into multiple tone area. Finally, records in each region of the image intensity extreme value and the gradient change direction are saved. By the way, the image intensity of rest position would be calculated by linear interpolation.
The experiments prove that limiting the amount of intensity change will affect the segmentation of watershed and the color reconstruction of linear interpolation. When the image intensity tendency is small variation, the fewer color composition regions are used to represent the image and the distortion is small, otherwise it needs more color composition regions to represent the image content.

並列關鍵字

Image Segmentation ； Image Representation ； Image Reconstruction ； Watershed Algorithm ； Asymmetric Image Representation ； Inclined Plane Decomposition ； Hierarchical Layer Representation

參考文獻

[2] L. Vincent and P. Soille, “Watersheds in digital spaces: an efficient algorithm based on immersion simulations,” IEEE Trans. Pattern Anal.MachineIntell., vol. 13, pp. 583–598, June 1991.

[3] R. Adams, and L. Bischof, “Seeded region growing,” IEEE Trans. Pattern Anal. Machine Intell., vol. 16, no. 6, pp. 641-647, June, 1994.

[4] Jaekyu Ha; Haralick, R.M.; Phillips, I.T., "Recursive X-Y cut using bounding boxes of connected components," Document Analysis and Recognition, 1995.,Proceedings of the Third International Conference on , vol.2, pp.952-955 vol.2,14-16 Aug 1995.

[5] J.Knipe and X. Li, “On the reconstruction of quadtree data,” IEEE Transactions on Visualization and Computer Graphics, Vol. 2, No. 2,1996, pp. 130-143.

[6] John M. Gauch, “Image Segmentation and Analysis viaMultiscale Gradient Watershed Hierarchies,”IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 8, NO. 1, JANUARY 1999.