My understanding is that, in very rough terms, the Tomasi-Kanade algorithm published in 1992 describes a way to reconstruct the 3D structure of an object from multiple images of that object, given that you have a reliable method of identifying important structural features from the 2D images.

A lack of general information on the topic has prompted me to ask this question: is this method deployed in any recent computer vision systems? Has it been followed up by other methods? Finally, if anyone would like to suggest any resources containing background information that are helpful in digesting the paper, please do. Despite being familiar with the mathematics, I had a hard time making any sense of it.


3D reconstruction method proposed by Tomasi and Kanade is a pioneering work and has delivered fruitful insights, but it does not seem to be commonly used 'as-is' in modern applications, or at least in research.

I guess this is because, at the moment of early 90's, there were not plenty numbers of local feature detector & descriptor as nowadays, ego-motion estimation (and 3D reconstruction) mainly relied on optical flow. However, since the appearance of local feature descriptors like SIFT, such ego-motion estimation method gradually shifted to utilize local feature descriptors to find sparse corresponding points among images due to its robustness and real-time performance. (Note that this is highly active topic in computer vision research, there are lots of proposals on how to estimate frame-by-frame motion. For your reference, one of the recent algorithms, named ORB-SLAM, integrates FAST as a detector and ORB as a descriptor.)

Talking about global structure, recent algorithms use bundle adjustment, which minimizes reprojection error to estimate motion in higher accuracy. This sometimes comes with pose graph optimization to handle loop detection in the sequence of images.

Sorry I have no idea which literature would help to read the article, but if you want to get a glimpse of what visual odometry is, I would suggest the tutorial coordinated by Prof. Scaramuzza.

Hope it helps.


Tomasi-Kanade factorization algorithm relies on the assumption of an orthographic camera model. Most of the current non-rigid SfM algorithms also rely on this assumption. So, although the algorithm in its original form might be outdated for ordinary SfM applications, some of the underlying ideas are still currently used in the non-rigid scenario.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.