Abstract: In order to solve the problem that the multi scale image noise smoothing easily lead to the loss of image edge details and important problem, a multidimensional space based on a step constrained noise image pre equalization algorithm is proposed. Analysis of the algorithm for noise image multiscale geometric space, will introduce multiscale transform domain denoising of nonlinear diffusion equations, and the multi-scale space noise suppression using pre equilibrium constrained nonlinear criterion, the denoising problem is transformed into constrained optimization solution of objective function, to maximize the noise interference on the edges, in the effective denoising and retain more edge information and details of images. The experimental results show that, in both the subjective visual effect and peak signal to noise ratio has proved the validity of the algorithm.