In most image processing processes that are utilized in most applications, denoising an image is a basic step. A two-phase conjugate gradient is used in our study to provide a method for reducing impulse noise. Noise candidates may be identified by an adaptive (center-weighted) median filter, which would then employ the conjugate gradient method to restore them. With this method, the noise candidates are recovered by reducing a functional for preserving edges. Its search strategy’s downward directionality is one of its most attractive features. The search for a global convergence of the Strong Wolfe line may turn up. Experiments with numbers show that conjugate gradient reduces impulse noise.