Abstract: Aiming at the problem that depth information acquisition method for the traditional moment-preserving universality is insufficient and edge is not ideal in practical applications, an improved moment-preserving depth information acquisition method is proposed. First, it gets the two images at different positions by changing the image distance and aperture parameters. Then it uses the method which is based on a combination of the image gradient and gray-scale standard deviation for detecting the edge of each image. Finally, according to the gradient image representing the ratio of the whole image, it calculates depth information. Experimental results show that the algorithm improves edge detection effect, reducing the amount of computation to improve the distance estimation accuracy and real-time.

Key words: moment-preserving, image distance, image gradient, gray-scale standard deviation

摘要： 针对传统矩不变深度信息获取方法在实际应用中普适性不足，且边缘检测效果不理想的问题，提出一种改进的矩不变深度信息获取方法。首先通过改变像距和光圈参数来获取不同位置的两幅图像，然后用基于图像梯度和灰度标准差相结合的方法检测出每幅图像的边缘，最后根据梯度图像占整幅图像的比值来计算深度信息。实验结果表明该算法提升了边缘检测的效果，减少了计算量，提高了距离估测的准确性和实时性。

关键词: 矩不变, 像距, 图像梯度, 灰度标准差