Abstract

Computer vision tasks often require the robust fit of a model to some  data. In a robust fit, two major steps should be taken: i) robustly estimate the  parameters of a model, and ii) differentiate inliers from outliers. We propose a  new estimator called Adaptive-Scale Residual Consensus (ASRC). ASRC  scores a model based on both the residuals of inliers and the corresponding  scale estimate determined by those inliers. ASRC is very robust to multiple- structural data containing a high percentage of outliers. Compared with  RANSAC, ASRC requires no pre-determined inlier threshold as it can  simultaneously estimate the parameters of a model and the scale of inliers  belonging to that model. Experiments show that ASRC has better robustness to  heavily corrupted data than other robust methods. Our experiments address two  important computer vision tasks: range image segmentation and fundamental  matrix calculation. However, the range of potential applications is much broader  than these.