Nicolau Werneck edited this page Sep 1, 2015 · 12 revisions
Clone this wiki locally

Corisco is a monocular vision method to estimate camera orientation. It is based on edgel extraction, and can analyze images with any camera model --- from images with ideal perspective projection or weak radial distortion to fisheye and equirectangular images. The M-estimation technique is used, and the optimization is carried out with the RANSAC and FilterSQP algorithms.

The doctorate thesis that originated Corisco is available for download, but only in Portuguese language [thesis page on USP system]. This Journal article, in English, also presents the technique: Corisco: Robust edgel-based orientation estimation for generic camera models, N. L. Werneck and A. H. R. Costa. Journal Image and Vision Computing 31(12), pp. 969-981, 2013. (also here)

(Observation: this project used to be hosted on code.google, but Github is now the official repo.)

How Corisco works (box diagram)


Here are some examples of Corisco applied to real images. The edgels are represented by small red line segments. They are all the data that is extracted from the image at the beginning of analysis. The result is displayed as the predicted orientations from the estimated solution, and they are generally aligned to the lines from the objects, and to directions of the extracted edgels.

Perspective projection

Radial distortion

Equirectangular (lat-lon)

Fisheye lens (polar azimuthal equidistant)