Abstract: In order to improve the positioning accuracy and deal noise of the multi-scale edge detection algorithm, a new method combines the force field transformation and multi-scale edge detection algorithm is proposed. New algorithm calculates inertial product energy of each pixel and uses it to obtain the force of pixels. In each scale, the image edge is obtained by self-adaptive threshold. The final image edge is received by executing from coarse to fine edge tracking. The experimental results show that the new algorithm effectively suppresses noise and improves the detection accuracy.

Key words: edge-detection, field transformation, inertial product energy, edge tracking, multi-scale

摘要： 为了提高多尺度边缘检测算法的定位精度，更好地抑制图像中的噪声和细节，将改进力场转换理论与多尺度图像边缘检测算法相结合，提出了改进力场转换理论的多尺度图像边缘检测算法。新算法计算每个尺度下图像中各像素点的内积能量值，利用内积能量值计算各个像素点间的引力；采用自适应阈值得到每一尺度下图像的边缘；在小尺度的检测结果下定位，执行由粗到细的边缘跟踪，得到最终的边缘结果。实验结果表明，新算法可以有效地抑制噪声，提高边缘检测的定位精度。

关键词: 边缘检测, 力场转换, 内积能量, 边缘跟踪, 多尺度