I'm looking for an Objective-C library / or just help in building a vision-based augmented reality application that does not rely on visual markers.
Qualcomm's is perfect, but only on Android (iOS is coming, but not soon enough). Any body know any other similar libraries?
QCAR for iOS has recently been released into public beta.
The only one I'm aware of is String, and I've become aware of that only via this iPhone + Kinect AR video (YouTube link) that recently did the rounds. So I've no direct experience of using it and no opinion on it, but I nevertheless think it qualifies as an answer to your question. It's commercial, but I think Qualcomm's is too.
Benjamin Loulier created a VRToolKit sample application for the iPhone that is based on the ARToolKitPlus framework. This sample uses the framework to track specific markers within the environment and allows you to overlay objects on them. I've seen some even more impressive demonstrations of this framework for using the camera to augment reality.
However, the ARToolKit framework (the parent of ARToolKitPlus) is by default available under the GPL license, which generally makes it incompatible with the App Store (and with commercial applications in general). You can purchase a commercial-use license for this framework from ARToolworks, if you need to use this within a non-GPL application.
Additionally, you might look into OpenCV for recognizing and tracking features in the environment. I've heard reports of people getting fairly decent performance out of it on iOS devices, even for more complex operations.
I've done some limited work myself in processing the video from the iPhone camera using the GPU, but nothing like what ARToolKit or OpenCV provides.
Have you looked at Aurasma http://www.aurasma.com/? They let you build custom apps with their system, which might suit your needs.
