Abstract: The threshold of Canny operator relays on the statistics of the image global gradient. The image global gradient differs from local gradient in most cases. Some obvious edge may be neglected while using the same threshold. Aiming at this problem and according to the image entropy, this paper proposes an edge detection algorithm for image partition and applies this algorithm in the Canny operator. Experimental results prove the validity of this method. Its detection effect is better than using Canny operator alone.

Key words: Canny operator, edge detection, entropy, image partition

摘要： Canny算子高低阈值的选取依赖图像全局梯度信息的统计，图像全局梯度信息和局部梯度信息的分布通常存在差异，导致在局部范围中较明显的边界可能被漏检。针对该问题，提出一种根据图像熵信息对图像进行分块的边缘检测改进算法，并将其用于Canny算子。实验结果证明了该方法的有效性，其检测效果优于直接运用Canny算子。

关键词: Canny算子, 边缘检测, 熵, 图像分块

CLC Number: 

• TN911.73