Abstract: Since classical morphological gradient operator cannot be directly extended to color image processing, the existing vector morphological gradient operator is sensitive to noise and it has low performance. This paper proposes a new soft vector morphological gradient operator and applies it in the edge detection of color images in order to solve the problem. New morphological gradient operator takes the advantage of soft morphology theory, which effectively reduces the influence of noise on the edge, so as to improve the anti-interference ability of the vector morphological gradient operator. Experimental results show that compared with the existing vector morphological gradient operators, the new soft vector morphology gradient operators achieve better gradient vector in the case of noise, and plays an important role in the following steps of color image segmentation and recognition.