An Analysis of ED Line Algorithm in Urban Street-View Dataset

Ciprian Orhei, Muguras Mocofan, Silviu Vert, Radu Vasiu

Abstract: Building detection in urban street view scenarios is becoming an important aspect of Computer Vision applications. In this paper we present an analysis of EDLine and Edge Drawing algorithms in a street-view dataset scenario when changing the first order derivative operator used inside the algorithms. To do so, we focused firstly on the general use case, using a natural image dataset, and secondly we looked on the effects we obtain on the use case of building detection in street view urban scenarios. We observed from our experiments that the proposed change brings marginal improvements to the algorithm that we present in the paper, visually and statistically.