Sparse pixel tracking: a fast vectorization algorithm applied to engineering drawings

We propose that the vectorization process be divided into two separate steps: basic vectorization, which yields coarse vectors, and post processing, which results in fine vectors. We propose shape (medial axis and line width) preservation and time efficiency as two criteria for evaluating basic vectorization. Inspired by the idea of the OZZ algorithm, we present a thinningless algorithm-sparse pixel tracking-for basic vectorization. Rather than visiting all the points along the medial axis of the wire´s black area, it sparsely visits selected medial points. The result is a crude polyline, which undergoes polygonal approximation to remove redundant points. Due to the sparse pixel examination and the use of a new data structure, the vectorization is highly time efficient, as well as accurate. Like OZZ, sparse pixel tracking preserves line width as well as the medial axis