کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
530098 | 869741 | 2015 | 11 صفحه PDF | دانلود رایگان |
• Balloon fitting (BF) is method to fit multiple connected ellipses to noisy figures.
• Connected ellipses are mapped to Gaussian mixture (GMM) and geometric constraints and then a modified constrained expectation maximisation (EM) is used to fit the ellipses to a silhouette.
• A Key modification of the EM algorithm is to increase precision each step (akin to blowing a balloon).
• BF and extensions have a superior root mean distance between silhouette boundary and ellipses compared to other methods of fitting multiple connected ellipses.
• Additionally BF decreases the chance of becoming stuck in an inferior local optimum.
This paper presents a framework to fit data to a model consisting of multiple connected ellipses. For each iteration of the fitting algorithm, the representation of the multiple ellipses is mapped to a Gaussian mixture model (GMM) and the connections are mapped to geometric constraints for the GMM. The fitting is a modified constrained expectation maximisation (EM) method on the GMM (maximising with respect to the ellipse parameters rather than Gaussian parameters). A key modification is that the precision of the chosen GMM is increased at each iteration. This is similar to slowly inflating a bunch of connected balloons and so this is called balloon fitting. Extensions of the framework to other constraints and possible pre-processing are also discussed. The superiority of balloon fitting is demonstrated through experiments on several silhouettes with noisy edges which compare other existing methods with balloon fitting and some of the extensions.
Journal: Pattern Recognition - Volume 48, Issue 7, July 2015, Pages 2198–2208