Abstract: A method of multi-scale edge detection for colour images based on the wavelet was studied. By selecting perceptible YUV colour space, the tensor product of multi-resolution-analysis was substituted for the tensor product of the wavelet function to build up the two dimensional Mallat algorithm to implement the fast wavelet transform, according to wavelet analytical theory. A standard pattern was applied to the multi-scale wavelet decomposition and the edge-detection results were obtained in the case of each scale. The experiment result coincided with theory analysis in high multi-scale and low multi-scale detection.