Title:	A NEW ALGORITHM FOR CONTENT-BASED 3D-IMAGE RETRIEVAL
Author(s):	Dumitru Dan Burdescu , Liana Stanescu , Razvan Tanasie , Anca Ion
ISBN:	978-972-8924-39-3
Editors:	António Palma dos Reis, Katherine Blashki and Yingcai Xiao (series editors:Piet Kommers, Pedro Isaías and Nian-Shing Chen)
Year:	2007
Edition:	Single
Keywords:	Content based region, object retrieval, 3D model, color.
Type:	Short Paper
First Page:	132
Last Page:	137
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	The image processing and the content-based image retrieval techniques are used to solve problems for different domains in the real world (medicine, robotics) where all the objects have three dimensions. The correspondence between the 3D object and the 2D image used in the applications is achieved by specific methods like projection. This transformation implies the loss of some information, the value of the third dimension being finally disregarded. For some purposes this loss might not prove significant, but to others it may be very important, and the final result might turn out to be wrong. In this paper we propose new and original algorithms for content-based 3D object retrieval. The 3D objects are firstly processed for obtaining 2D slices, so that each 3D object is represented by a set of significant 2D slices. The distance between the slices is chosen equal to a virtual spatial network edge length, thus the number of slices depend on the third dimension of the object. These slices are further used for content based region query on color feature. The image slice pixels are arranged into small hexagons (like honeycombs), so each image slice is viewed as a graph not as a pixel matrix. The vertices represent the pixels and the edges represents neighborhood between pixels. An image slice is segmented into color regions characterized by color and area. The user can retrieved the scenes that contain objects similar with query objects of interest. The designed new algorithm has a better time complexity than other similar algorithms because it does not compare the 3D objects voxel by voxel [7], instead, it only compares the voxels (from the query and current image) situated in the nodes of the virtual spatial network, thus significantly reducing the computation effort.
	Go Back