PPT Image Segmentation Based on Color Entropy

Image segmentation is a preliminary requirement to hybrid image encoding. In this paper, we propose SIG, an algorithm based on information gain for slide image segmentation. SIG algorithm depends on two features observed from the slide image samples: color entropy distribution and information gain....

Full description

Saved in:
Bibliographic Details
Published in2015 IEEE Trustcom/BigDataSE/ISPA Vol. 2; pp. 147 - 152
Main Authors Feijuan He, Xuelu Chen, Dawei Xu, Xianglin Miao
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2015
Subjects
Algorithm design and analysis
Color Entropy
Entropy
Gray-scale
Hybrid Image Coding
Image coding
Image color analysis
Image segmentation
Information Gain
Segmentation
Transform coding
Online AccessGet full text
DOI10.1109/Trustcom.2015.574

Cover

More Information
Summary:Image segmentation is a preliminary requirement to hybrid image encoding. In this paper, we propose SIG, an algorithm based on information gain for slide image segmentation. SIG algorithm depends on two features observed from the slide image samples: color entropy distribution and information gain. SIG algorithm consists of two stages. First, it finds the seed of monochrome, textual and grayscale images following the greedy strategy, and then it expands the seed image to generate image blocks seeking maximum information gain. Empirical experiments indicate that SIG-based hybrid image encoding is superior to JPEG2000 as well as 8×8 and 16×16 block-based hybrid image encoding in terms of efficiency.
DOI:10.1109/Trustcom.2015.574