Multi-Resolution and JND-Based Image Segmentation Algorithm

• Main Authors: Ming-Tzung Huang, 洪銘聰
• Other Authors: Day-Fann Shen
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/84233388135855627981

碩士 === 國立雲林科技大學 === 電機工程系碩士班 === 90 === Image segmentation is a basic step of image processing , such as computer vision , image analysis, medical image and the object-oriented compression standard : MPEG4. The complexity of image segmentation is a barrier of image real-time applications nowadays. Because the segmentation results are judged by people, a good image segmentation algorithm should utilize human perception properties effectively. In the thesis , our algorithm utilizing the JND (Just Noticeable Different) and multi-resolution properties can reduce computation load effectively and get a concise result. Besides, we also proposed another image segmentation algorithm without watershed processing , we call it “the JND Pyramid Image Segmentation Algorithm”. It is a hybrid of edge detection and region growing techniques ,and it can produce the different segmentation levels to represent an image effectively.