Locally Optimized SSAC: LOSSAC [ ]


A new method of locally optimized sample consensus based on S-estimator has been introduced in this paper (LOSSAC). Locally optimized RANSAC (LORANSAC) and other locally optimized procedure are developed based on M estimator. The weakness of M estimation is the lack of consideration of distribution of data and it is not a function of overall data, because it uses median as the weight value. The proposed method uses the residuals standard deviation to overcome the weaknesses of median. The performance of the improved SSAC is studied and compared with LORANSAC and LOMSAC. The LOSSAC provides reliable results and more efficiency by comparing with the number of inliers estimated under various procedures.