Jan’s Experiments

So I finally tried to implement my own automatic way of stitching panos. Don’t expect some sophisticated system, it’s actually pretty easy. I still have to click my mouse going from one step into another, however the point is that this can be fully automated and it did not require any sophisticated input. A chimp monkey could do it instead of me . I spent about 30 seconds with creating this pano which is far less than the time it took me to put this this blog post online.

Here’s what I did:
- Take the pictures, there are two things that needs to be considered when taking pictures for automatic panos
- 1. Leveling: The horizon needs to be leveled as precise as possible because the automatic software cannot determine the correct leveling for you
- 2. Precission: The pano head parallax point needs to be precisely calibrated, any misaligned is more visible in auto mode
- Load the pictures into a computer
- Adjust chromatic aberation and via a predefined filter plus apply any other filter you wish
- Export them as tiffs
- Load them in PtGui and let them auto-stitch, there is one important point:
- 1. Lens settings in PtGui: I don’t have the numbers, I’m running completely on auto stitch, that’s why I need even better precision when taking panos
- Save the result as jpeg

That’s it. 30 seconds of manual input plus 2 minutes of my quad-core gaming machine time.

And here’s the result:

Tram 7 in Prague

What do you think? I know the quality could be better, I saved it in a very low resolution to improve load time and like I mentioned above, it would help to have a lens profile set correctly in PtGui and/or calibrate the pano head more precisely.

Google Street View is a feature of Google Maps and Google Earth which provides panoramic views of streets so the viewers can look around at any place that has been covered. What is the difference between Google Street View and handcrafted virtual panoramic photography as of today? What is stopping Street View to look as today’s best VR panoramas? What are the current and future technology constraints? How will Street View look in 5 years?