A Layered Depth Image (LDI) is one of the popular representation and rendering methods for 3D objects with complex geometries. In this paper, we propose the new compression algorithm for depth information of a 3D object represented by LDI. For the purpose, we introduce the concept of partial surfaces to seek highly correlated depth data irrespective of their layer and derive a depth compression algorithm by using them. Partial surfaces are approximated by a Bézier patch and residual information is encoded by a shape-adaptive transform. Experimental results show that our proposed compression method achieves a better compression performance than any other previous methods.
INTRODUCTIONImage-based modeling and rendering techniques have received much attention as a powerful alternative to the traditional geometry-based techniques for generating novel views of real/synthetic scenes or objects. Lots of useful geometry-based modeling and rendering techniques have been developed. However, they require elaborate modeling and long processing time to reconstruct and render complex objects. And they provide unsatisfactory quality in producing truly photorealistic objects. As an attractive alternative to overcome these problems, many of image-based modeling and rendering (IBR) techniques have been proposed in the past decade. They use 2D images as primitives to generate an arbitrary view of the 3D scene [19]. Among the variety of IBR techniques, a layered depth image (LDI) [1] is one of the efficient rendering methods for 3D objects. LDI represents the object with an array of pixels viewed from a single reference camera. As a data structure corresponding to a single projection of scenes, it stores information for each intersection point of each projection ray, which start at a fiducial point, with objects. And a layered depth pixel stores a set of depth pixels along one line of sight sorted in front to back order. The front pixel in the layered depth pixels samples the first surface seen along that line of sight; the next pixel samples the next surface seen along that line of sight, etc. Although LDI is one of the efficient rendering methods, one big obstacle is that it has a huge amount of data. In addition, because of multiple layer characteristic of LDI, the number of layers (NOL) must be additionally encoded.There are many of lossy 2D image compression algorithms and schemes. However, because of the special data structure of LDI, they cannot be applied directly or are not very efficient. So, several researches have been carried on the compression of LDI: compression for real world scenes from natural multiview video [4][5][6][7][8][16], lossy compression for synthetic static scenes [9][11], lossy compression for static objects [10]. Previous approaches have focused on layer-based coding like Fig. 1(a): Most of them, such as DA [2][3][5][6][7][9][11][16], MD [8], NFFL [2][3][5][6][7], and NFNL [4], treat each layer of LDI as a 2D image. And some few of them, such as CMP [10], generate new 2D images and focuses on how to re...