LAS vs LAZ: What Is the Difference?

LAS is the binary exchange format for point clouds, maintained by the ASPRS LAS Working Group. A file starts with a header — point count, bounds, the scale and offset of the coordinates, and records that carry the coordinate system — followed by one record per point.

Each point stores X, Y and Z as integers that the header's scale and offset turn back into coordinates, plus attributes such as intensity, return number and classification; depending on the point format it also carries GPS time and RGB colour. Classification is what lets software separate ground (class 2) from vegetation and buildings.

LAZ is a LAS file compressed with LASzip, an open compression developed by the lidar pioneer Martin Isenburg and published as an Apache 2.0 library. It is lossless: decompressing a LAZ gives back every bit of the original LAS, so coordinates, classes and colours are identical. The LASzip project puts a LAZ at 7–20% of the size of the LAS.

Nothing about the data changes: same points, same attributes, same coordinate system. What changes is the cost — a LAZ is several times smaller to store and to send, and it is decompressed as it is read, which software does on the fly.

A COPC (Cloud Optimized Point Cloud) file is a LAZ 1.4 file whose points are stored in a clustered octree. An ordinary LAZ reader still reads it from start to finish; a reader that understands the index can fetch only the region and level of detail it needs, which is what makes COPC suited to serving very large clouds over the web.

Keep and exchange LAZ: it is the same data at a fraction of the size, and most point cloud and terrain software reads it. Use LAS only for a tool that cannot read LAZ, and go back to LAZ afterwards — the conversion is lossless in both directions.

Before choosing, check the things that matter more than the extension: is the coordinate system written in the file, are the points ground-classified, and is there RGB colour? Those decide what you can make from the cloud.

STREAM opens LAS and LAZ directly; a LAZ is not unpacked to LAS first. The import dialog reads the cloud before you commit — point count, classification, RGB and bounds — and disables what the file cannot give: without ground classes there is no bare-earth DEM, without colour no orthophoto, and the dialog says why.

The importer streams the points in fixed-size chunks and turns them into a DEM, a surface model or an orthophoto at a resolution suggested from the point density. STREAM does not display or edit individual points; it builds the terrain you measure and design on.