图像分类识别是计算机视觉系统的重要组成部分,而正确快速地提取目标特征参数是图像分类识别的前提.本文针对目前图像几何特征参数提取算法存在运算量大的问题,提出一种快速的几何特征提取算法.算法首先采用顶点链码对图像连通区域进行编码,使之映射为一个闭合区域,然后将顶点链码的方向码与离散格林定理相结合,把曲面积分变为曲线积分,使多种几何特征参数的计算均可统一在该算法框架下.实验表明,该方法具有多参数计算统一性、运算量小、可适用于特殊区域等特点.%Image identification and classification play an important role in the field of computer vision and extracting the features quickly and correctly is the essential precondition for image identification and classification. To solve the problem that the algorithm of extracting geometric features has large amount of calculations, this paper proposes a faster algorithm. The algorithm firstly encodes the image connected region by using the Vertex Chain Code, making it mapped as a close area, then turns the surface integral into line integral by combining Vertex Chain Code with Discrete Green, unifying the calculation of the various parameters under the framework of this algorithm. The experimental results show that this algorithm has the features of the unity of calculated parameters, fewer computations and applicability for special area.
展开▼
