Imagine IG v3 ready for production usage

Innovation has been a major part of Imagine DNA since the company has been founded in 2005. Thanks to a high image quality and blazing performance, the second version of our Image Generator software has been successfully set up in more than 20 sites around the world.

Now, since the Image Generator market is very competitive, since technology constantly offers more opportunities, we decided to put a lot of efforts in making our new generation of Image Generator software a great leap forward, compared to the previous version.

We consider IG v3 is now mature enough and ready for production use. As it is still based on Microsoft Direct3D 10, no hardware change is required.

See below the highlights of this new release:

• Hi-resolution terrain with geo-morphing and dynamic streaming
  While the v2 terrain was completely static, with a good resolution around the ARP, but a much poorer quality once the observer moves away from the center of the scene, the new terrain offers a much higher texture definition, constant across the whole scene, wherever the camera is. 
  This is achieved by dynamically streaming and paging the relevant chunks of the terrain, according to the position of the camera.
  Also, the terrain features built-in geo-morphing capabilities, which results in no visible popping effect when terrain LODs changes occur.
  
  
• Support for geo-referenced imagery
  V2 customer scenes were mainly relying on Google Earth screenshots, which were used to texture the terrain. While being a simple and handy workflow, this source of imagery is not optimal when image quality is concerned:
  
  • Google Earth picture is using textures tiles which have been heavily JPEG compressed (to reduce network bandwidth ), resulting in a blocky and pretty low-quality source image.
  • Since the IG has to submit the textures to the GPU in a compressed DXT form (in order to offer real-time performance), two different lossy compressions scheme are being applied consecutively on the textures, which, of course, is not great.
  • Google Earth pictures are already re-projected in Googles reference system. A projection is a kind of rotate/scale/shear filter applied to the image, resulting in a resolution loss.
  • Of course, as the images are not geo-referenced, the customer using several satellite/aerial images has to manually scale, translate (and sometimes rotates) them, to have them aligned together (and match the optional underlay CAD file).
  Hence, using geo-reference satellite imagery solves all these issues:
  • no successive lossy compression schemes applied anymore
  • no successive projections applied anymore
  • automatic relative placement of all geo-referenced images used

• Unlimited number of real lights with a low performance overhead
  We have rewritten a substantial part of the renderer. While still based on D3D10, and therefore compatible with the hardware used for v2, we switched to a much more modern technique called “deferred rendering”. This technique allows the usage of hundreds of real light (point and spot lights) for a very minor performance overhead.
  In v3, all the runway, edge and taxiway lights emit real light, which means, when an aircraft rolls on the taxiway, it gets lit by all these small lights (including specular), resulting in a much more realistic image.
  Also, all the aircraft lights (navigation light, taxi lights, strobe lights, anti-collision lights, etc…) also emit real lights.

• Betters shadows
  We have improved our shadow implementation, allowing a bigger shadowed area and resulting in no more visible seams. Moreover, the terrain is able to cast shadow as well, resulting in fantastic results in mountainous areas.

• Betters physics
  We have improved the physics performance and stability, this is mostly noticeable in ADT.

• Much faster generation of the airport 3D model in FAB
  We have split the generation process into two parts:
  • Generation of the terrain (using height map and satellite imagery)
  • Generation of the vector layer (runways, taxi ways, roads, etc) and 3D objects
  We made this choice as we think that the first part (terrain) doesn’t change that often, therefore, it’s not necessary to generate it every time. It has to be generated once, or when the user wants to add/modify the satellite imagery. Only the second part changes frequently. Also, we removed the PRT pre-processing step, which results in the 3D-model integration process being one order of magnitude faster.

• Backward-compatibility with FAB v2
  The v3-based FAB offers backwards compatibility with FAB v2 files. This means, all you’ll have to do to make your scenes work in this new v3 version of the IG, is open the v2 airport in FAB v3, and generate the airport model again. That’s it.

• Much faster loading time (order of magnitude faster)

• ...and despite all these new features: better performance

Now, as images are always better than words, see below a side-by-side comparison of v2 and v3.

Below a trailer of this new release: