« Google Licensing Automated 3D Building Generator | Main | New Layer for German Google Earth »

20 de Mayo 2007

Virtual Earth 3D Comparisons with Google Earth

Since this blog is obviously focused on Google Earth, I have not devoted a great deal of attention to Microsoft's new competing product: Virtual Earth and, specifically, Virtual Earth 3D (VE3D). However, Microsoft has been making significant progress with their application (still in beta) and especially their data, so I feel I should at least share some of my thoughts and observations in comparison to Google Earth. I will be sharing some video, screenshots and a table of comparisons in this article. Please take note my bias is towards Google Earth (GE), but I've tried to be fair in my comparisons. And also, the focus of this comparison is from a user perspective, not a developer.

First off, I want to say that Microsoft recently began supporting the 3DConnexion SpaceNavigator. This greatly improves the experience with VE3D. The VE3D mouse interface is a bit clunky in comparison to Google Earth (you can't seem to spin the globe for example). But, with SpaceNavigator you get the same smooth 3D movement you have been able to get with GE since last November. There are some differences in the visuals - not just based on the data. More on this later. In the meantime, here is a video showing VE3D with the SpaceNavigator in use:

Note: I've made some comments in the video about the quality of the renderings and buildings. Here is a my original video review using SpaceNavigator with GE. There are more detailed comparisons below...

Here's a quick summary of the more significant things I've seen Microsoft add to VE3D since the beta was released last fall:

  • SpaceNavigator support
  • Support for the Firefox browser in addition to Internet Explorer for VE3D
  • Huge amounts of new data. Aerial photography, and Bird's Eye views for many new cities in Europe and the US, about 30 more full cities with 3D building data, and more cities with road layers.

Although Microsoft still does not have nearly the amount of coverage of the world with high resolution aerial photos as Google, the amount of data they have added is significant. The Bird's Eye views in particular add several times as much photography for a given city as the standard straight-down views. So, for any given city where Microsoft has Bird's Eye views they have more aerial data than Google in those cases. And, Microsoft is adding cities with this data at an amazing rate. Especially when you consider that it costs many millions of dollars to fly, photograph, process, store, and deliver this amount of data.

The 3D building data sets are also very impressive. Being able to automatically generate 3D buildings has some advatages and disadvantages (see some earlier comments and video I've written about). The advantages are being able to have a vast majority of buildings modeled, and be able to generate more cities with this level of detail at a better cost per city. The disadvantage is that the level of detail and accuracy (at least currently shown) in VE3D is lower quality when examined closely.

Here are some screenshots of 3D buildings for comparison between Virtual Earth and Google Earth. I realize some of you will say I'm not making fair comparisons here, but read on, I try to summarize my comparisons fairly in the conclusions. Click on the thumbnails to see full details. First we compare the Denver, Colorado capitol building in Virtual Earth (autogenerated) to the hand-built building in Google Earth. This example illustrates some points.

Virtual Earth 3D Denver Capitol Building   Google Earth 3D Denver Capitol Building

Next we have the same building with the cityscape in the background and closer-up view of the building. Note that the auto-generated textures in VE3D do not look very good up close. Meanwhile, there is a great deal more detail in the hand-made GE model (especially if you zoom in close).

Virtual Earth 3D Denver Capitol Building 2   Google Earth 3D Denver Capitol Building 2

Part of the problem with the VE3D view is due to Microsoft's decision to not load too many texture details because they take up a lot of memory. On older machines, Google's approach will greatly overload the system causing much slower performance. However, on newer machines, Google's performance actually is faster even with a heavily detailed number of building models like found in Denver. But, there is merit to Microsoft's decision to have lower detail since a larger number of users will have fewer problems. It is also important because Microsoft has nearly 50 cities with large numbers of 3D building data, and is adding several more every month or so.

Here is a comparison of a hand-built model of the US Capitol building in VE3D verses the hand-built one in GE. Again, Microsoft has chosen much less model complexity, but the model looks quite acceptable.

Virtual Earth 3D US Capitol Building   Google Earth 3D US Capitol Building

Here's a quick table of comparisons between VE3D and GE - (again, I'll add the caveat I am a big Google Earth fan):

Virtual Earth
Google Earth
VE3D is much more usable now that it supports SpaceNavigator
GE is also better with a SpaceNavigator
When rendering it's aerial and satellite photos, Microsoft uses a tiling approach which is very annoying to view as you zoom, pan and fly along. You can see "chunks" of imagery load and unload as you move along.
GE's graphics engine has been refined over the past several years, and the algorithms and technology seem more pleasing to the eye. Transitions between imagery is more smooth.
Microsoft licensed the rights from a company which has a patented method for taking not only normal straight-down aerial photography, but also oblique Bird's Eye views. And, Microsoft acquired a company whose cameras are ideal for this application. [Corrected based on comment.] When they capture a city, they have a top-down view - plus FOUR views from the North, East, South, and West. The Bird's Eye views are very high resolution and are a great resource for different perspectives. These photos are also used by Microsoft to auto-generate 3D buildings for a city. The Bird's Eye views are a major asset, but they are not as well utilized because the concept, and user interface, I think is difficult for the average user to understand.
Google has no real equivalent to the Bird's Eye views.
Microsoft's technology for automatically generating 3D buildings is a decided advantage over Google's approach for capturing entire cities in 3D - at least today. Microsoft has already released about 50 cities with 3D building data. This is definitely goodness. The amount of 3D Building data is impressive in scale, and quite decent in quality until you get closer to the buildings.
Google has been asking the community at large to build 3D buildings and share them through their 3D Warehouse. Thousands of buildings have been created all over the globe and the models eventually can become a part of GE's 3D Building layer. This approach seems to be working to a degree, but only a small percentage of a typical city will get built typically using this approach. Most people aren't going to want to take the time to render every building on every block. Also, the quality of the models varies. Google is planning to announce they are licensing technology which will enable them to automatically generate 3D buildings at Where 2.0. In this area, Google will be playing catchup. But, they will have the advantage of both approaches once the new technology and data is in place.
Microsoft's decision to support lower resolution textures is good from the perspective performance and better support for people with slower/older machines.
Google's higher quality buildings look better when viewed up close, but at a potential performance penalty if you have a slower/older machine. If you have a new machine, GE will typically give you the same or better performance and the better quality buildings.
Virtual Earth does not have many other kinds of data layers. They have the "Maps" and "Hybrid" layers which show roads an road label information and borders and country names. And you can load other "collections" of information from other people. But, there are not a lot of other kinds of layers. They also have some tagged information for points of interests like restaurants, but not wide spread information in this area.
Google has a huge number of built-in layers for things like roads and road labels, and borders as well. But, Google goes MUCH further with all kinds of points of interest layers: Airports, railways, ferry routes, restaurants, lodging, etc. Also, data from a number of organizations like National Geographic Magazine, Discovery Channel, UNEP, AIA, and much more.

That's all the time I have for now. If you are a serious Google Earth user, you might still want to give Virtual Earth 3D a try. Just be aware that you have to have Windows XP SP2 or Windows Vista to be able to use VE3D. It doesn't work on the Mac OS X or Linux. By the way, to do a really cool direct comparison of these two products, check out Globe Glider. The 3D buildings collections Microsoft has done are very impressive. And, the Bird's Eye views with the VE2D application are definitely worth exploring as well. For pretty much everything else, at this point, I would recommend Google Earth. 'Nuff said.

Enviado por FrankTaylor at 20 de Mayo 2007 a las 02:36 PM

  • Google Earth Blog © 2005, 2006, 2007 Copyright de Frank Taylor. Todos los Derechos Reservados.
  • Todas las imágenes de Google Earth capturadas de pantalla son Copyright de Google

  • Comentarios

    At least you know you are biased :-)

    Microsoft has not acquired Pictometry (the people who have the technology for Birds Eye imagery), but they have signed a deal with them for US, and with Blom (who has the rights for it in Europe) for Europe.
    They acquired Vexcel however, a company who makes digital metric cameras for mapping purposes / top down imagery. This is now known as "Microsoft Photogrammetry". You can read more about the "UltraCAM" cameras if you Google for it. The newest one is no less than 218 megapixel!

    Enviado por: Morten at 20 de Mayo 2007 a las 09:53 PM

    I'm quite impressed by how far VE has come along. It's really quite slick, especially with the SpaceNavigator. Two things still really bug me about it:

    1 - No loading status. I don't know if it's done loading data in my current view or not. GE makes that quite obvious.

    2 - Image labels as raster instead of vector, which causes most of the "chunk" issues that Frank was referring to. If the roads/cities/etc were vector graphics separate from the images that would smooth things out a lot. Not as much as GE, but at least a little closer.

    Enviado por: Mickey at 20 de Mayo 2007 a las 11:27 PM

    Frank, you know a little more about my history, but I'll make myself perfectly clear about why it is that I am also biased toward Google Earth.

    In the big picture between both apps, Google Earth has always been ahead of VE in relation to end-user navigation. This is primarily due to what was started as Keyhole, but I think the idea for Microsoft was mismanaged more on the idea that they would tie it in with the Xbox controller, which ultimately alienates the average user (if you're going to go this route - then why not support the PS2/3 controller - or at the very least ensure elegant useability for the average end-user?).

    That being said - when you even attempt to compare the two using the Connexion controller, there still runs a major faux pas in the comparison. Google Earth's navigation applications are still superior to VE in various ways. Meaning, for the common, everyday end-user who's just navigating through using VE or GE with the typical mouse controller - they'll gain a much smoother and user-intuative experience using GE.

    I think you've generated a very fair highlight of the differences using the controller, however - and it does appear to be rather smooth. But when you break the two applications down to their lowest common demoninators, I still can't help but retain my bias.

    I'm rather critical as well about the quality of the models in VE, and prefer GE's more hand-touched models. The only complaint that I might have regarding the GE models, is that I think the textures could possibly be shaded to match the histograms of the imagery beneath. That would create more perceptual realism without loss of such custom quality. (For me, perception within the environment matters. It's the difference between something looking realistic, and non-realistic or 'cartoony'.)

    I also haven't purchased the Connexion controller (yet), so I suppose one could say that I'm also biased because I include myself as one of the users I mention - which is more likely the majority of the demographic of people using these applications.

    Enviado por: Daniel at 21 de Mayo 2007 a las 07:00 AM

    I agree with Daniel about the building shading. Google's models are more detailed but still look cartoonish and out-of-place because the color and brightness do not match the underlying imagery. At a distance, the vector edges often overwhelm the models; check out Sutro Tower in San Francisco which looks almost black because of the lines.

    Microsoft's buildings are better matched tonally to their surroundings and to each other, and for this reason I think they look better from a distance. And the fact that they have modeled even the minor buildings really makes the cities come alive.

    Hopefully google's new technology will rectify this, because otherwise their product is far superior.

    Enviado por: Lauren Steely at 26 de Mayo 2007 a las 01:43 AM

    Envíe un Comentario:

    NOTA: Use español o Inglés. Los Comentarios son revisados previa publicación.