Painterly rendering using image salience
Painterly rendering using image salience
复制标题
DOI：
10.1109/eguk.2002.1011281
复制
发表时间：
2002-06
期刊：
影响因子：
--
通讯作者：
John Collomosse;Peter Hall
中科院分区：
文献类型：
--
作者：
John Collomosse;Peter Hall
本文的贡献是一种新颖的非真实感渲染（NPR）技术，能够在2D图像（如照片）上产生人工“手绘”效果。我们的方法不需要用户交互，并利用图像显著性和梯度信息来确定单个笔触的隐式顺序和属性。我们的技术的好处是完全自动化，并减轻在绘画过程中图像细节的损失。来自图像较低显著区域的笔画不会侵犯较高显著区域；这可能发生在一些现有的绘画方法中。我们详细描述了我们的算法，并用图库说明了它的应用。
The contribution of this paper is a novel non-photorealistic rendering (NPR) technique, capable of producing an artificial 'hand-painted' effect on 2D images, such as photographs. Our method requires no user interaction, and makes use of image salience and gradient information to determine the implicit ordering and attributes of individual brush strokes. The benefits of our technique are complete automation, and mitigation against the loss of image detail during painting. Strokes from lower salience regions of the image do not encroach upon higher salience regions; this can occur with some existing painting methods. We describe our algorithm in detail, and illustrate its application with a gallery of images.