Randomized RANSAC with Td,d test

摘要: Abstract Many computer vision algorithms include a robust estimation step where model parameters are computed from data set containing significant proportion of outliers. The ransac algorithm is possibly the most widely used estimator in field vision. In paper we show that under broad range conditions, efficiency significantly improved if its hypothesis evaluation randomized . A new (hypothesis evaluation) version algorithm, r-ransac , introduced. Computational savings achieved by typically evaluating only fraction points for models contaminated with idea implemented two-step procedure. mathematically tractable class statistical preverification test samples For this derive an approximate relation optimal setting single parameter. proposed pre-test evaluated on both synthetic and real-world problems increase speed shown.