Lidar depth image compression using clustering, re-indexing, and JPEG2000

Large LiDAR (Light Detection And Ranging) data sets are used to create depth mapping of objects and geographic areas. The suitability of image compression methods for these large LiDAR data sets was explored, analyzed and optimized. Our research interprets LiDAR data as intensity based "depth images", and uses k-means clustering, reindexing and JPEG2000 to compress the data. The first step in our method applies the k-means clustering algorithm to an intensity image creating a small index table, an index map and residual image. Next we use methods from previous research to re-index the index map to optimize compression when using JPEG2000. And lastly we compress both the reindexed map and residual image using JPEG2000, exploring the use of both lossless and lossy compression. Experimental results show that in general we can compress data to 23% of the original size losslessly and even further allowing for small amounts of loss.