摘要 |
PURPOSE: A gradual vocabulary tree structure building method for image data search is provided to ease implementation by applying a depth map including brightness information to a vocabulary tree structure. CONSTITUTION: Local feature information is extracted by inputting a depth map of an object unit and a gradual vocabulary tree is composed by using the local feature information(S100). An excellent node is selected by using feature information of a leaf node of the gradual vocabulary tree(S200). The gradual vocabulary tree is updated by inserting the local feature information into the leaf node(S300). If the gradual vocabulary tree is updated, the excellent node is reselected in a previous step(S400). [Reference numerals] (AA) Start; (S100) Extract feature information of an image and build an initial gradual vocabulary tree; (S200) Select superior nodes of the gradual vocabulary tree; (S300) Extract feature information and search a corresponding leaf node when a new image is inserted; (S400) Update the gradual vocabulary tree
|