一种结合边缘与区域信息的图像特征提取算法

An image feature extraction algorithm based on edge and regional information

摘要:

为了能够较准确地表示图像的形状特征，提出了一种融合边界和区域信息的新的形状描述子。首先对图像进行二维离散余弦变换，获得低频系数作为区域特征。之后提取图像的轮廓并进行采样，形成描述形状轮廓的有序点列表，对每个采样点分别顺时针和逆时针等距离跟踪获得2个邻点，计算拱高以及质心距离。然后获取由拱高和质心距离组成的复函数的频域描述子，并组合区域特征与轮廓特征。对MPEG-7标准图形库的检索实验显示，该描述子的检索性能显著优于三角形面积函数、质心距离函数与拱高半径复函数等同类描述子。

Abstract:

A new shape descriptor with fusion contour and region information is proposed in order to represent the shape characteristics of the image more accurately. Apply two-dimensional Discrete Cosine Transform(DCT) to obtain the low-frequency coefficients as regional characteristics. Then extract and sample the contours of image, and get the ordered list of points. Track each sampling point equidistantly clockwise and counterclockwise to get two adjacent points, and calculate the arch-height and centroid distance. Construct a complex function of arch-height and centroid distance, and obtain the DCT coefficients, then combine regional and contour features. Retrieval experiments on MPEG-7 shape database show that, the retrieval performance of this descriptor is better than that of the triangle area function, the centroid distance function, and arch-height radius complex function significantly.