A New Spatial Encoding Method for Sparse Binary Data

Abstract

Spatial encoding is an important technique for spatial binary data storage, which differentiates the defined data from the undefined data in many computer graphics applications. However, previous methods make significant memory cost when the spatial data becomes very sparse, particularly for high resolution domains. In this paper, we present a new sparse encoding method which has less storage requirement. It is simple and efficient in storing spatial binary data. Our method utilizes commonly used array to store the specific dimensional coordinate for each defined data. To achieve efficient access on the GPUs, we recur to an auxiliary table to record the first defined element of each row. Compared to several popular approaches, the experimental results on 2D images and 3D models demonstrate that our method has a better memory performance on spatial binary data.