Abstract: The essence of wavelet transform profilometry is to calculate the position of the wavelet ridge, and then obtain the optimal scale and phase information. However, the extraction of wavelet ridge can be easily affected by noise, which can result in the inaccuracy of wavelet ridge extraction. Aiming at this problem, a new cost function is proposed to extract the wavelet ridge. The cost function is established by using wavelet transform coefficient amplitude and scale parameter curve gradient information. The feature scaling method is used to adjust wavelet transform coefficient amplitude and scale parameter curve gradient to balance the weight of the two indexes. The wavelet ridge will be affected more along with the noise increase. The spatial noise information is combined to compensate the adjusted scale parameter curve gradient, increase the weight of scale parameter curve gradient, and make the wavelet ridge smoother. The experiment shows that the algorithm has good anti-noise ability and robustness, and can extract the wavelet ridge accurately.