一种用于图像检索的DP匹配方法

Dynamic Programming Matching Method for Image Retrieval

相似性测量是图像检索中的一个重要步骤,距离度量是相似性测量的一个重要方法,DP匹配是距离度量的一种特殊形式。该文在分析一维DP匹配的基本原理后,提出用能量矩阵代替DP匹配中的距离矩阵。能量矩阵是根据图像分割区域的边界点的能量来定义的。图像检索实验结果表明,改进的DP匹配方法在图像检索中的应用是可行的,且有一定的优越性。

Similarity measurement is an important step in image retrieval. In the general retrieval, distance measurement is the common method, and DP matching is one special form of distance measurement, using matching distance matrix. This paper analyzes the basic theory of DP matching, uses matching energy matrix instead of matching distance matrix, thereby gains the similarity of the query, image and the sample image. It gains the character energy matching matrix according to the defined confine-points＇ energies, Based on this algorithm and a conceived prototype, experimental result proves the algorithms is feasible.