Abstract: Aiming at detection challenges of contaminated DataMatrix (DM) code coexisted with perspective distortion, in this paper we presents an extracting of best edges combination approach based on Hough transform domain in which linear characteristic is embodied.Firstly, marginal points' transformation from image to Hough domain is performed fleetly using principle direction of DM code.Afterwards, we are able to obtain the valid combination of candidate marginal points after prior rulesbased filtering of non-marginal points.Finally, a distribution model of inside edges of DM code in multi-perspective is build, thus extracting features, and the edges combination with largest eigenvalue is estimated as an ultimate result.The experimental results demonstrate the greater robustness and flexibility of our proposed approach to accurate detection of Data Matrix with contamination and perspective distortion, compared with traditional methods.