Application of Information Measure in Segmentation of Hydrophobicity of Insulators
Bing Qi, Liangrui Tang, Jing Zhang · 2008
In order to effectively distinguish object from background, three essential characters of edge-points are taken into account, which are consistence, direction and degree information in neighbor pixels. According to these features, firstly, three information measures are proposed to quantitatively depict corresponding characters, which can be represented by gray values, gradient values and similarity degree in neighbor pixels. Secondly, experimental results signify that all the characters have effects on extracting object from background respectively and also show which character has larger influence on extracting droplets (or watermarks) from hydrophobic images. Finally, these three information measures are pieced together to form the feature domain for FCM clustering, which are applied in segmentation of hydrophobic images. Experimental results show that the proposed algorithm has great application in segmentation of hydrophobic images, for it's efficiency of extracting shape information of droplets (or watermarks).