摘要
计算机图形分离就是在由计算机生成的图形和自然图像合成的混合图像中将人工区域和自然区域分开 利用自适应自组织映射对彩色图像进行量化 ,针对每种颜色组成一个色平面 ;然后将色平面转换成二值图像 ,对其中的连通区域进行快速标记 ,并提取出边缘 将标记后的区域和边缘映射到原图像 ,定义并计算每个区域的粗糙度和边缘对比度 ,最终完成各个连通区域的识别 实验结果表明 。
Aiming at differentiating computer graphics from natural image in a color image, color quantization is achieved by adaptive self-organizing maps and the number of colors is reduced to facilitate the subsequent segmentation and recognition of color images. Color planes are formed by extracting every kind of color and then transformed into bi-level image, where connected-components are labeled and their edges are detected. Afterwards the labeled connected-components and their edges are mapped back into the original color image to calculate inhomogeneity and edge contrast. Generally, inhomogeneity of artificial regions and edge contrast of natural regions are low. According to the ratio of inhomogeneity and edge contrast, all regions can be classified. Some experimental results verified the presented method.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2004年第8期1040-1044,共5页
Journal of Computer-Aided Design & Computer Graphics
基金
国家创新研究群体科学基金 ( 60 0 2 43 0 1)资助