Loading...
Point localization in multi-camera setups has been widely studied in computer vision. Recently, in finite-resolution camera settings, a consistent and optimal point localization algorithm called SHARP has been proposed, under the assumption of noiseless camera poses and error-free matching. In this work, we relax this assumption on noiseless camera poses and propose a new point localization algorithm. We formulate this point localization task as a gradient-ascent optimization function, for maximizing the objective function under computational geometric constraints. Experimental results verify the efficacy of our approach as compared to the current state-of-the-art localization algorithms.
Use this identifier to reference this record
Type
report
Authors
Publication date
2015
Total of pages
68
Note
Doctoral internship report
EPFL units
Available on Infoscience
February 19, 2016