Color Image Segmentation with Multi-level Semantic Framework
WU Ruimina, TV Arts · Video Engineering · 2011
A new method to segment color image with a multi-level semantic framework is presented in this paper.Through user interaction,the objects in multi-regions are abstracted,and their sub-objects in different semantic levels can be recognized and the semantic relationships between any two of them are established.The semantic relationships feed back to the segmentation and makes the results better.The experiment is performed and presented on different kinds of images and some frames of several video sequences.The segmentation results of every level and final semantic hierarchy of objects are looking forward to work effectively in image and video editing based on objects or parts of objects.