A novel algorithm for achieving Wavelet transform on the Cellular Neural Network Universal Machine (CNN-UM) visual neuroprocessor is presented in this work. The CNN-UM is implemented on a VLSI programmable chip having real time and supercomputer power. This neurocomputer is a large scale nonlinear analog circuit made of a massive aggregate of regularly spaced neurons which communicate with each other only through their nearest neighbors. VLSI implementation of this circuit is feasible due to its locally connectivity and fixed output function of each cell consisting of a piece-wise linear saturation function imposed by the difficulty of realizing non-linearities in hardware. In the next, implementation of wavelet transforms by means of an analog algorithm is presented. Thus, we can use the CNN-UM in solving realtime applications where wavelet are an essential step like computer-vision algorithms for stereo vision, binocular vergence control, texture segmentation and face recognition. .