Better Color Coordinates for Consistent Color Image Segmentation

• Main Authors: Ming-Cheng Kan, 甘敏成
• Other Authors: Chung J. Kuo
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/68168919036144288730

碩士 === 國立中正大學 === 電機工程研究所 === 89 === Color image segmentation is achieved through direct combination of the segmented image from color component images. Since many color coordinates are available, there must be one that is most suitable for image segmentation. This thesis proposed agreement, orthogonality and spectral similarity measure as criteria to decide which coordinate is the best. According to our simulation results, is always the best coordinate for color image segmentation no matter which measure is employed. The organization of this thesis is as follows. Chapter 2 reviews the watershed algorithm used for image segmentation and some backgrounds about color coordinates. The criteria and procedures to decide which color coordinate is the best is shown in Chapter 3 and Chapter 4 is experimental results. Finally the conclusion and discussions are given in Chapter 5.