基于改进Canny算子的手势图像边缘检测

针对手势检测不准确,手势边缘检测有偏差等问题,文中提出了一种改进的Canny算子用于检测手势图像边缘。改进的Canny算子使用自适应中值滤波与双边滤波相结合的方法去除图像噪声,并在Sobel的基础上扩展45°和135°方向的梯度幅值计算,自适应获取最佳阈值判别图像边缘。最后,使用哈夫变换与像素点的梯度方向相结合检测边缘,连接边缘、增强边缘。实验表明,改进后Canny算子检测的结果8连通数比传统Canny算子检测的结果8连通数多,边缘连续性较好,检测精度更高且自适应性增强。

Gesture Image Edge Detection Based on Improved Canny Operator

Aim at the problem that the existed gesture detection was inaccurate and the gesture edge detection was deviated, an improved Canny operator was proposed to detect the edge of the gesture image. The improved Canny operator used adaptive median filtering combined with bilateral filtering to remove image noise. On the basis of Sobel, the gradient amplitude calculation in the 45° and 135° directions was extended, and the optimal threshold was adaptively obtained to discriminate the image edge. Finally, the Huff transform combined with the gradient direction of the pixel was used to detect, connect and enhanced the edge. Experiments showed that the results of the improved Canny operator detection were more connected than the traditional Canny operator. The edge continuity was better, the detection accuracy was higher, and the adaptability was enhanced