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. 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 8x8 and 16x16 block-based hybrid image encoding in terms of efficiency.