Title

An investigation into the use of jpeg image compression for digital photogrammetry: does the compression of images affect measurement accuracy

Document Type

Conference Paper

Rights

This item is available under a Creative Commons License for non-commercial use only

Publication Details

European Conference on Geographic Information Systems (EGIS) Proceedings; Paris, France; March, 1994

Abstract

Combining techniques used in image processing, digital photogrammetry has been proven to significantly increase productivity over analytical methods. In certain areas, digital photogrammetry has shown itself to be far superior to conventional methods of data collection. Softcopy (digital) photogrammetry raster files take up large amounts of storage space. Larger pixel sizes may help to reduce file size, but it has been shown in studies that larger pixel sizes may lead to less accurate results. What is the user to do? The importance of file compression should not be underestimated. File sizes affect almost every step of the digital photogrammetric workflow. Without some type of significant file compression, the entire process becomes complicated with excessive data file management. Depending on the specific workflow, file sizes are also critical in desktop environments where data storage may be small (e.g. heads-up digitising on digital orthophotos in a PC environment). The JPEG image compression/decompression algorithm is a compression method which utilises a variable compression factor or Q-factor. Files can be significantly reduced in size without any visual loss of image quality. Typical reduction is one-third to one-fourth the uncompressed size. The purpose of this paper is to compare the accuracy and speed of processing uncompressed digital imagery with that of compressed imagery throughout a digital photogrammetric workflow. Each process was timed, including; film scanning, file transfer, epipolar resampling, automatic DTM collection, orthophoto production, creating image pyramids and image backup and restore. By using only hardware and software to perform the above tasks, human operator bias will be removed from the process completely.