the Art of Geography weblog

Microsoft Virtual Earth adds Birds-Eye View to 3D Globe: “

We all somewhat guess this would be coming around and tonight Microsoft releases Birds-eye View for their 3D Globe. They’ve taken those great Birds-eye View images from the 2D maps and using Photosynth they stiched them all together to create a ‘3D World’. Now of course this isn’t a true 3D, but if you’ve ever tried out Photosynth, you’ll know what I’m talking about. The Virtual Earth Blog spells it out quite nicely.

For background, its important to understand the challenges of visualizing our Birds eye imagery in a seamless mosaic the way we are all used to looking at satellite imagery that looks straight down at earth. Since all of the images are shot from the same point of view, it’s relatively easy to stitch them together in a convincing tapestry. There’s still challenges like doing good color balancing across images and rectifying so that buildings in tall cities don’t appear to butt heads, but these are pretty well understood problems. Birds eye images are a different story. because of the way they are captured, there is no easy way to stitch them at their edges without introducing nasty distortions. The result is that Birds eye imagery is viewed as discrete ’scenes’ instead of 1 giant tapestry. when you navigate to the edge of the current scene, the most appropriate next scene is dynamically determined, then displayed. Since Birds eye imagery is captured from 4 angles, we have North, South, East and west views of each point on earth adding another dimension of complexity to navigation.

In 3D, the same challenges remain, but are easier to address with client side managed code. the result is a user interface that borrows concepts from MSR’s Photosysnth to allow you to navigate more fluidly and naturally than what is possible with a javascript based client.

…

The street level images are nice from a navigating sense, but I always feel like I’m in a canyon. These Birds-eye view images are much more open and having them available in 2D or 3D really pushes Virtual Earth ahead of Google Maps in my book (now how about making the darn think work in Safari?). One thing that jumped out at me was how good Microsoft’s building models look. Sometimes you can’t tell the difference between the Birds-eye and the 3D model. For those who can’t install the 3D add on, take a look at the video below.

(Via Spatially Adjusted with James Fee.)

Carnegie Mellon University researchers, working with NASA Ames Research Center scientists, have developed an inexpensive robotic device that allows any digital camera to take gigapixel panoramic photographs, known as GigaPans. The technology is being used by students to document their communities and by the Commonwealth of Pennsylvania to make Civil War sites accessible on the Web. The system uses a tripod-like mount to allow digital cameras to take hundreds of overlapping images of landscapes, buildings, or rooms. Software developed by Carnegie Mellon and Ames is used to arrange the images in a grid and digitally merge them together to create a single image that could contain tens of billions of pixels. Carnegie Mellon has also created a Web site so users can upload and interactively explore the panoramic images in any format. “An ordinary photo makes it possible to cross language barriers,” says Illah Nourbakhsh, an associate professor in the School of Computer Science’s Robotics Institute. “But a GigaPan provides so much information that it leads to conversations between the person who took the panoramas and the people who are exploring it and discovering new details.” Nourbakhsh hopes that GigaPan will help develop a community of producers and users. “GigaPan is not just about the vision of the person who makes the image,” Nourbakhsh says. “People who explore the image can make discoveries and gain insights in ways that may be just as important.”

Full article is at http://www.cmu.edu/news/archive/2007/September/sept26_gigapan.shtml