A method for picture processing is provided. The method includes the following. Multiple pictures are obtained through photographing. At least two pictures are selected from the multiple pictures, and according to facial information of each of the at least two pictures, feature information of a preset facial part in each of the at least two pictures is obtained. One or more target pictures are determined in response to the feature information of the preset facial part in one of the at least two pictures indicating that the preset facial part is in a preset feature state, where the one of the at least two pictures is determined as one target picture. When multiple target pictures are obtained, a multi-frame denoising processing is performed on the target pictures to obtain an output picture.