Abstract: There are a lot of arc light noise and spatter point noise in the images using the CCD to capture CO₂ arc welding pool. It is difficult to detecte accurately the edge of the images on the melting pool. Median filter and multiplication are used to reduce the noise of the CO₂ arc welding image. Basing on the zero crossing and Trous algorithm, we have researched into a detecting method and its algorithm by using the Bubble function, compared and analyzed the influence of different scales wavelet on the effect of the edge detection. The result shows that there is a better effect by using Bubble function to construct wavelet base, and this found a good basis for controlling the melting depth, the seam tracking and so on.