基於色彩強化與量化之彩色影像分割

COLOR IMAGE SEGMENTATION BASED ON COLOR ENHANCEMENT AND COLOR QUANTIZATION

葉修銘(Hsiu-ming Yeh)

指導教授 ： 周俊賢

大同大學/電機資訊學院/電機工程學系所/碩士(2005年)

摘要

為了改善影像的品質，數位影像強化的技術因此被發展出來。一般來說，影像強化的技術一般應用在RGB色彩模型上，因為不需複雜的演算法則即能輕易快速的進行處理。
在之前的研究中，有一種在向量空間上增強特定方向向量的方法稱為向量擴大法，這是藉由增加彩色向量的強度用以靠近重要或感興趣的色彩（COI），來達到影像強化的目的。此外，向量擴大法需要複雜的運算，由於RGB色彩空間並非均勻空間，所以強化後的結果並不是令人非常滿意。
在均勻的色彩空間上如Lab上，可直接以Euclidean計算所得差異用以代表視覺上之色彩差異，方便強化之用。因此，基於向量擴大法的觀念，在均勻色彩空間內減少Euclidean距離差異可達到影像強化的目的。
另一個有趣的議題為如何產生影像分割所需要的COI。藉由向量量化器由原始影像產生色彩叢集，並將色彩叢集內相近的顏色在色彩差異之區域內進行合併，以產生具代表性的COI。因此影像分割可藉由反覆的色彩強化及量化，在均勻色彩空間內達成而不造成色彩資訊的溢位及失真。

關鍵字

色彩強化 ； 彩色影像分割

並列摘要

In order to improve the quality of imagery, digital image enhancement techniques have been developed. In general, the image enhancement is applied in the RGB color space that can be implemented simply and effectively without computationally intensive.
In previous research is to increase the directional resolution using a concept that calls Vector Amplification, increasing the magnitude of color vectors close to the direction of the color of interest (COI) to enhance image. The vector amplification may need more computationally intensive. Furthermore, the image enhancement doesn’t have qualified result in RGB color space.
Due to advantages of color differences in uniform color space such as Lab, Euclidean distance has close correlation with the perceptual color difference in the uniform color space. Therefore, based on the algorithm of Vector Amplification, the image can be enhanced by reducing the Euclidean distance with specified multi-color of interest in uniform color space.
Another interesting issue is how to generate COIs for image segmentation. First, cluster of colors are generated by the color quantization from the original image. Then representational COIs are generated by merging color element of cluster within the region of color difference. Hence, image can be segmented by iterating color enhancement and color quantization in uniform color space without any overflow color information and color distortion.

並列關鍵字

color enhancement ； color image segmentation

參考文獻

[1]S. J. Sangwine, B. N. Gatsheni, T. A. Ell, “Vector amplification for color-dependent image filtering,” ICIP (2), pp. 129-132, 2003.

[4]J. Fan, D. KY Yau, A.K. Elmagarmid, W. G. Aref, “Automatic image segmentation by integrating color-edge extraction and seeded region growing,” IEEE T IMAGE PROCESS 10 (10), pp. 1454-1466 ,Oct. 2001.

[5]A. Martínez-Usó, F. Pla, P. García-Sevilla, “Color Image Segmentation Using Energy Minimization on a Quadtree Representation,” ICIAR (1), pp. 25-32, 2004.

[9]S. Makrogiannis, G. Economou, S. Fotopoulos, N.G. Bourbakis, “Segmentation of Color Images Using Multiscale Clustering and Graph Theoretic Region Synthesis,” SMC-A(35), No. 2, pp. 224-238, Mar. 2005.

[2]N. Le Bihan and S. J. Sangwine, “Quaternion principal component analysis for color images,” IEEE International Conference on Image Processing (ICIP 2003), Barcelona, Spain, pp. 809–812, 14-17 Sept. 2003.