A VLSI architecture for 2D mesh-based video object motion estimation

This paper proposes a novel 2D mesh-based object motion estimation architecture that generates a content-based video object. The video object is represented using an alpha plane and a texture. The alpha plane consists of two layers (1) a mesh layer and (2) a binary layer that defines the object boundary. The texture defines the object's colors. The mesh layer is represented by a new adaptive hierarchical mesh. The proposed mesh is formed by recursive triangulation of the initial coarse mesh geometry, which reduces the mesh code. The proposed architecture generates a progressive mesh code. The proposed architecture offers low power performance at the algorithm level as follows: (1) decomposing the mesh into independent triangular patches that can be processed in parallel, (2) using a hierarchical structured triangulation technique, that simplify the processing unit, and (3) using the three steps motion estimation algorithm to simplify the motion estimation of the mesh nodes. The proposed architecture is suitable for very low bit rate online applications and it can be used as a building block for an MPEG-4 codec.