Structural-context-preserving image abstraction by using space-filling curve based on minimum spanning tree

We propose a novel image abstraction method which generates a non-photographic painting-like image automatically from a photographic natural image while preserving its structural context, detailed structure, and so on. In the proposed method, concretely, one-dimensional vector ε-filter is applied to an input image along a space-filling curve. This space-filling curve is constructed by using a minimum spanning tree reflecting structural context of the input image. This novel image abstraction approach is completely different from conventional approaches. By using the proposed method, aggregation of texture details and enhancement of sharp edges are effectively realized while preserving structural contexts such as thin lines, sharp corners, and so on. The structural-context-preserving abstraction capability of the proposed method is verified by some experimental results and examples.