Pen-Ink Rendering for Traditional Building Images

Abstract

This investigation proposes a non-photo realistic rendering to reflect enough to emphasize the delicate pen touch of Korea traditional building images. The earlier studies for pen-ink sketch technologies produce good results for the general images but they should not make successful results for the images of traditional architectures. It because that the image comprises a large proportion of repetitive patterns of exquisite lines and curves in its roof and latticed door. For featuring out the distinguishable characteristics of the traditional building images, the important thing is to decide the direction and length of strokes. We suggest a method to extract stroke’s information using Tensor Subspace Analysis (TSA) technique and to draw pen-ink strokes according to the information. Therefore, the proposed method generates pen-ink sketches reflecting the texture property for the traditional building images and emphasizing the characteristics of light and shade for each area.