SEPTEMBER 2001

8

Chromo-Stereoscopy: 3D Stereo -Use the 3D ChromaDepth Viewing Spectacle Included in

A very modern trend in photogrammetry is the automated production of DEM data from

overlapping stereo-pairs of remotely sensed image data, followed by the generation of

an orthoimage or an orthoimage mosaic. These operations may be carried out as a

single integrated process using suitable imagery acquired from a spaceborne platform.

By Gordon Petrie, Thierry Toutin, Hasin Rammali & Christophe Lanchon

Examples are (i) the overlapping imagesacquired cross-track from the SPOT or IRSsatellites on two different orbits; and (ii)those acquired along-track from a singlepass of the MOMS, IKONOS or EROS-A1satellites. (iii) A further possibility is to useeither stereo- or interferometric radar (SAR)imagery to generate the DEM and orthoim-age data. Even more common is the use ofimagery taken from an airborne platform.Within this latter context, the process ofDEM and orthoimage production is com-monly carried out by mainstream pho-togrammetric service providers using stereo-pairs of aerial frame photographs, whethercollected in analogue or digital form. Nowthe method is being implemented with thedigital imagery acquired by the new genera-tion of airborne pushbroom scanners, e.g.the DLR HRSC-A and LH Systems ADS40imagers. The great attraction of the processfrom the photogrammetric production pointof view is the highly automated procedurethat can be implemented with a minimumof human input after the identification andmeasurement of the ground control points(GCPs). While this is extremely satisfactoryfrom the point of view of the photogram-metric service providers, the situation maybe viewed in a quite different way by theactual clients or end-users.

Need for 3D Stereo-Viewing

On the one hand, the client or end-userdoes receive an ortho-rectified image hav-ing a good map geometry, together withthe corresponding DEM data. On the otherhand, delivery of an orthoimage throws thetask of carrying out the interpretation andextraction of the point, line and area dataon to the end-user. Previously, when thetraditional type of vector line map with con-

tours was being supplied to the client, thedetection, identification and extraction ofthe topographic features was carried out asa prior operation by the photogrammetristand the specialist image-interpreter using3D stereo-viewing. Nowadays the end-usermay have to execute this task himself usingthe orthoimage without the benefits of hav-ing a 3D stereo-model available for the pur-pose. Besides which, 3D stereo-viewing isalso essential for the execution of the quali-ty control that is a necessity with DEMsproduced using automated image matchingmethods. This quality control operationtakes the form of editing and removing mis-matches and ensuring that the DEM dataand the lakes, rivers, mountain crests, pass-es and ridge lines occurring in the area ofinterest are all consistent with one anotherand make sense from the topographic point

of view. Quite apart from feature extractionand DEM quality control, the simple stereo-viewing of an area in 3D is often highlybeneficial to an understanding of the terrainand its landforms and hydrology by fieldscientists, engineers, planners and militarypersonnel. However there is no possibilityof them doing so using only the DEM andorthoimage data.

3D Chromo-Stereoscopic Images

One solution to this particular problem isto generate a 3D chromo-stereoscopicimage from the orthoimage and DEM data.This method of 3D stereo-viewing has beendeveloped originally by Steenblink in thelate 1980s. During the 1990s, it thenbecame available from the ChromatekTechnologies company in the U.S.A. as acommercial product, called ChromaDepth.This has been further developed andrefined specifically for use with remotesensing data by Dr. Toutin of the CanadaCentre for Remote Sensing (CCRS), whichforms part of Natural Resources Canada(NRCan). Details of the underlying theoryand the actual process of generating chro-mo-stereoscopic images are given in two ofhis papers (Toutin 1995, 1997).

(a) The refraction of white light through a glass prism; (b) The superchromatic double prism; (c) ChromaDepthviewing spectacles using a pair of double prisms (drawn by Mike Shand)

Chromo-Stereoscopy: 3D Stereo

SEPTEMBER 2001Latest News? Visit www.geoinformatics.com

The Basis of Chromo-Stereoscopy

The basic process is relatively straightfor-ward - at least in principle! The individualelevation values from the DEM are encodedsystematically into the orthoimage on apixel-by-pixel basis in the form of differentcolours. This allows an observer to exploitthe well-known phenomenon that objectslying in the same plane - e.g. on the screenof a display monitor or on a hard-copy print- but in a different colour, appear to anobserver to lie at different depths. Thus ared object appears to the observer to lieabove the level of a blue object. While ayellow object appears to lie at an interme-diate level or depth between the red andblue objects. To further enhance this effect,use is made of suitably designed spectaclesthat utilize the differential refraction of lightthat is produced by prisms. These prismscause differential refraction of the variouscomponents of white light. These spreadout to produce the classic "rainbow" effectin which the light occurring at differentwavelengths is spread out and seen as con-tiguous bands of different colours. Utilizingthis effect, the ChromaDepth spectacles that

appearing to lie nearer (i.e. higher) to theobserver, while those encoded in blue (withthe lowest elevation values) appear to liefurther away. Terrain objects having interme-diate elevation values between these twoextremes (which have been encoded inorange, yellow, green, etc.) will appear atthe appropriate intermediate depths or lev-els between these highest and lowest ele-vation values.

Examples

The chromo-stereoscopic technique can beused with a wide range of imageryacquired from a variety of spaceborne andairborne platforms; using different types ofimaging sensors; and with images havingwidely differing ground resolution values.Three examples covering quite differenttypes of terrain at a variety of image scalesand ground resolutions are given below.

-Montreal, Quebec, CanadaThis image is a space radar orthoimagemosaic covering the city of Montreal and

are worn by the users are equipped withpairs of specially manufactured prisms.These are used to enhance the apparentdepth of the differently coloured parts ofthe image for each eye and so produce a3D stereoscopic image.

ChromaDepth 3D Stereo-Viewing

The ChromaDepth spectacles utilize a pairof very thin prisms made of a completelytransparent plastic material. In practice,these act like thick glass prisms. To achievethis effect, each prism comprises two com-ponent thin prisms cemented together - theone using a low dispersion material, theother a high dispersion material. The com-bined effect of the pair of these doubleprisms when used in the spectacles is todecode the colour-encoded orthoimage asseen by each eye. This allows the observerto experience an increased perception ofdepth corresponding to the elevation valuesthat are present in the DEM. Thus a strongstereoscopic impression of the terrain isobtained with those objects having thehighest elevation values (encoded in red)

with Orthoimages & DEM Datathis Magazine!-with Orthoimages & DEM Data

Figure 1: Montreal, Quebec Canada (copyright: Canadian space Agency)

9

SEPTEMBER 2001

10

its surroundings. It has been produced byCCRS/NRCan from two Radarsat-1 SAR finemode images having a 6m ground pixelsize. The chromo-stereoscopic mosaic hasbeen generated using a two-stage process.(i) First of all, each of the radar images hasbeen fitted to GCPs extracted from thestandard 1:50,000 scale topographic mapsof the area produced originally by theCanada Centre for Topographic Information(CCTI/NRCan). This was followed by theactual ortho-rectification process using theDEM data generated by CCRS/NRCan fromthe digitized 10m contour lines taken fromthe topographic maps. The merging of thetwo individual ortho-rectified images pro-duced the required radar orthoimage mosa-ic. (ii) The second stage in the process wasto integrate this radar orthoimage datawith the DEM data to generate the finalchromo-stereoscopic image. This was car-ried out via a radiometric process involvingthe use of the IHS (intensity-hue-satura-tion) transformation to define and generatethe appropriate colours for each individualpixel in the radar orthoimage mosaic. Thiswas done on the basis of the correspond-ing elevation values derived from the DEM.The final 3D chromo-stereoscopic mosaicwas published originally by CCRS/NRCan inthe form of a poster, with the actual imagemeasuring 53 x 35cm. However it has ofcourse been reduced greatly in size andresolution for inclusion in GeoInformatics.The image shows the St. Lawrence Rivercrossing the area from the lower left cornerto the top right corner of the image. TheOttawa River flows into it as a tributarystarting midway up the left side of the

image. The city of Montreal occupies thecentral part of the image with the MontRoyal featuring prominently above the St.Lawrence River.

-Misratah, Tripolitania, LibyaThe second example forms part of thecoastal area around the city of Misratahlocated 150km east of Tripoli, the capital ofLibya. The area has been used for the pro-duction of a series of experimental spaceimage maps intended to explore the carto-graphic design aspects of this type of mapat 1:50,000 and 1:250,000 scales. Theunderlying objective behind this project wasto see if this type of image map would be asuitable replacement for (i) the existingseries of line maps @ 1:50,000 scale thathad been produced for the more developedareas of Libya; and (ii) for the series ofspace image mosaics @ 1:250,000 scale thatprovide complete national coverage of thecountry. Both series are now completely out-of-date. As part of this experimental workcarried out at the University of Glasgow,chromo-stereoscopic images of the test areawere produced with the help of Dr. Toutin.On this occasion, the data that was usedwas SPOT Pan monoscopic imagery with a10m ground pixel size. As with the previousCanadian example, the required DEM wasformed from the digitized contours takenfrom the existing 1:50,000 scale topographicmap. The procedure used for the generationof both the SPOT orthoimage and the finalchromo-stereoscopic image was basically thesame as that used with the Montreal exam-ple. The final image shows clearly thestrongly incised wadis (valleys) cut into the

semi-arid plateau and the slopes runningdown to the flat coastal plain whereMisratah city and the developed areas ofagricultural land are located.

-Sophia Antipolis, Southern FranceThe third example shows part of the "sciencecity" of Sophia Antipolis which lies a shortdistance inland from the Cote D'Azur and justnorth-west of the coastal city of Nice in thesouth of France. This image has been gener-ated by the ISTAR company, whose headquar-ters and main processing facilities are actual-ly located in Sophia Antipolis. The 3Dchromo-stereoscopic image has been pro-duced from overlapping high-resolutionimages (with 0.25 ground pixel size) of thearea acquired by DLR's HRSC-A airbornepushbroom scanner. This features a three-linepushbroom arrangement with the overlappingimages generated from the forward, nadirand backward pointing linear arrays. Thisallows stereo-coverage to be collected along-track during a single flight. The HRSC-Aimager has its own integrated DGPS/INS sys-tem which provides continuous positionaland attitude information for each line of theimage data. In this particular case, first of all,the elevation data was generated from theoverlapping HRSC-A stereo-imagery usingautomatic image matching techniques in con-junction with the in-flight position and atti-tude data given by the DGPS/INS unit. Thistook the form of a digital surface model(DSM) including both building and treecanopy elevation data. Thereafter the ortho-image was produced using the DSM data asthe basis for the ortho-rectification. This was

Figure 2: Misratah, Tripolitania, Libya (copyright: University of Glasgow)

SEPTEMBER 2001Latest News? Visit www.geoinformatics.com

followed by the production of the final chro-mo-stereoscopic image (called a"Hypsoimage" by ISTAR) using the sametechnique as described for the two previousexamples. Once again, the pronounced depthof the 3D image at the large scale of theimage shows clearly the rolling wooded char-acter of the area and the heights of the manylarge buildings that have been constructed bythe industrial, commercial and scientificorganisations that have established them-selves in Sophia Antipolis.

Conclusion

The combination of orthoimage and DEMdata - when allied to the complementary

Canada, 588 Booth Street, Ottawa, Ontario,

K1A 0Y7, Canada

Dr. H. Rammali (hrammali@hotmail.com) Surveying

Department of Libya, Algeria Square, Tripoli. Libya

Mr. C. Lanchon (lanchon@istar.fr) Marketing

Department, ISTAR, 2600 Route des Cretes, 06905

Sophia Antipolis, France

References

Chromatek Inc., Alpharetta, Georgia, U.S.A. - Web

Site:- http://www.chromatek.com/

Th. Toutin & B. Rivard, 1995 - A New Tool for Depth

Perception of Multi-Source Data, PE&RS, Vol. 61, No. 2:

1209-1211.

Th. Toutin, 1997 - Qualitative Aspects of Chromo-

Stereoscopy for Depth Perception, PE&RS, Vol.63, No.

2: 193-203. ■

chromo-stereoscopic image that can bederived from these two data sets - is wellsuited to a wide range of topographic map-ping, GIS and thematic mapping applica-tions where 3D stereo-viewing would behelpful, indeed often invaluable, to theusers. As the examples show, the techniquecan be applied to a wide range of space-borne and airborne imagery having quitedifferent ground resolutions.

Professor G. Petrie (g.petrie@geog.gla.ac.uk)

Department of Geography & Topographic Science,

University of Glasgow, Glasgow, G12 8QQ,

Scotland, U.K.

Dr. Th. Toutin (thierry.toutin@ccrs.nrcan.gc.ca)

Canada Centre for Remote Sensing, Natural Resources

Figure 3: Sophia Antipolis, Southern France (copyright: ISTAR)

11