一种图像粒标记模型及其实现

An Image Granule Labeling Model and Its Implementation

针对图像分析领域缺乏可扩展的基础模型,对灰度图和彩图的标记模型展开研究。通过分析粗糙集和商空间理论的适用性,综合图像标记处理的特定应用需求,引出概念粒和连通粒2个概念,构建粒标记模型。基于通常情况下图像尤其是彩图标记中粒的数量巨大、结构复杂等现状,定义行连通段、潜在连通范围,引入动态预设标记集,简化连通判定,给出实现粒标记模型的线性算法,即图像粒标记算法。用二值图和彩图分别作验证和对比分析,结果表明,该标记算法有效、精确,且较传统标记算法更高速。

With regard to the absence of base models which are extendible in the field of image analysis,an image granule labeling model is proposed. On the analysis of the applicability of rough sets and quotient space,two new concepts,semantic granule and conception granule,are defined and the image labeling model based on granules is constructed with them. A linear algorithm,Image Granule Labeling（IGL）algorithm,is presented for the realization of the granule labeling model with the definitions of connected segment in a line and potential connection range between two lines. It simplifies connectivity determination using a dynamic set of provisional labels and takes into account the fact that the quantity of granules may be huge and the structure of granules is very complex normally. Comparisons and experimental results on binary images and color images show that the proposed granule labeling algorithm is effective and accurate,and it is quicker than conventional labeling algorithms.