[4H2-GS-4] Knowledge utilization and sharing: Society and knowledge

[4H2-GS-4-05] Application of Hierarchical Entropy to Image

〇Nao Uehara¹, Teruaki Hayashi¹, Yukio Ohsawa¹ (1. The University of Tokyo)

Keywords:Image, Segmentation, Entropy, Recognition

We calculate the entropy of the image in which the areas of it are divided based on Efficient Graph-Based Image Segmentation.The change of the entropy is calculated by changing the size of the combined pixels(component)and the size of the rectangular areas(mesh) in the image.The relationship between the change of the entropy and the structure is verified.It is concluded that the entropy decreases with increasing the size of components.It is also found that the change of the entropy of the characteristic part of image is different from the other part.In addition, it is considered that the information of an important part of image is lost before and after the change of the entropy decreases drastically.