优化分割的手绘图像彩色化技术

Hand-Drawn Image Colorization Based on Optimized Segmentation

图像彩色化技术是计算机图形学与数字图像处理中重要的研究课题,目前该技术主要基于笔触交互方式完成,所以精确的着色区域分割是提高彩色化质量的关键.手绘黑白图像往往有复杂的线条、纹理和不封闭的轮廓,对分割造成了很大的困难.为解决这一问题,提出一种交互式彩色化方法.首先运用一个高质量、高效率的边缘保持滤波和拉普拉斯算子来解决灰度不连续的问题,强化区域边界信息,并计算灰度能量;然后利用标色信息计算标色能量,克服用户不精确标色的问题;最后将灰度能量和标色能量构建成一个能量函数,利用Graph Cut技术检测最优分割.该方法不受内部纹理、边界缺口和不精确标色的影响,可以更准确地划分着色区域.实验结果表明,由于获得的分割信息更准确,文中方法对于多种手绘图像能够取得更好的着色效果.

Image colorization is an important topic in the field of computer graphics and image processing. Accurate image segmentation is usually vital for the colorization results in interactive methods. However, in hand-drawn images, complicated lines and patterns as well as discontinuous outlines always make segmentation a hard problem. We propose a method to solve the problem and segment related regions more precisely. Firstly, we use an edge-preserving filter and a I.aplace edge detector to smooth color discontinuities and strengthen outlines. Secondly, grayscale energies are computed. Next, stroke energies are calculated according to positions of user strokes to overcome imprecise placements. Finally, grayscale and stroke information are modeled into an energy function, and an optimal segmentation is obtained based on Graph Cut theory. Experimental results show that this method, with more precise segmentation results, can obtain high accurate results for various kinds of hand-drawn images.