Effect of Data Compression on Dem Accuracy Derived from Satellite Images

Very little work has been done on the systematic evaluation of effect of compression on the accuracy of DEM derived using satellites imageries. Most of studies have been carried out for aerial images and models have been developed which relate DEM accuracy with compression. One such model predicts effect of compression on DEM accuracy as 0.2% of flight height for CR up to 15:1 but these models cannot be used for space-borne data because 0.2% DEM accuracy as predicted by the model may cause only a few meters error in height for air borne data, whereas, in case of space borne data, it becomes few kilometers. Here we have studied the impact of compression on DEM accuracy derived from space borne sensor data and results have been recorded. For this purpose, we have used two compression techniques : namely JPEG compression and wavelet based set partitioning in hierarchical trees (SPIHT) coding to study the effect of compression on the accuracy of DEM. The results include impact of compression on DEM in terms of statistical errors measured on the point to point basis for DEM grid of 150x150 size and visual quality scores. We have used 'Softspace' photogrammetry software package developed at Dept. of Space, India , which generates DEM and ortho-images from stereo pair acquired from Indian Remote Satellites IRS 1C/1D and SPOT. We observe from the results that there is no noticeable degradation on DEM accuracy up to the compression ratio of 25:1. It is also observed that statistical errors in DEM for JPEG at a given compression ratio is better than that for SPIHT but visual quality of DEM is better with SPIHT compressed data.