ECISER: Efficient Clip-art Image SEgmentation by Re-rasterization

摘要: Abstract Clip-art image segmentation is widely used as an essential step to solve many vision problems such colorization and vectorization. Many of these applications not only demand accurate results, but also have little tolerance for time cost, which leads the main challenge this kind segmentation. However, most existing techniques are found sufficient purpose due either their high computation cost or low accuracy. To address issues, we propose a novel approach, ECISER, well-suited in context. The basic idea ECISER take advantage particular nature cartoon images connect with aliased rasterization. Based on relationship, clip-art can be quickly segmented into regions by re-rasterization original several other computationally efficient developed paper. Experimental results show that our method achieves dramatic computational speedups over current state-of-the-art approaches, while preserving almost same quality results.