The invention discloses an image layer cutting method based on weighting cut combination. The method comprises the following steps: step 1, constructing a weighted graph in a mode that a pixel of an image is taken as a top point, and constructing a cost function of weighting cut combination of an undirected weighted graph; step 2, selecting a seed point, defining a sparse interpolation matrix, updating and combining weighting values of the types represented by the seed point, carrying out layer iteration so as to obtain an optimized similar matrix, roughening the original image in different layers and stopping the iteration till obtaining an outstanding interest region, and solving a characteristic system represented by an optimized cost function so as to obtain the rough cut of the outstanding interest region in the undirected weighted graph; and step 3, scanning the boundary from the top to bottom aiming at the outstanding interest region, and combining the boundary by using an inversed interpolation method so as to obtain the optimal cut of the image. A normalized image cutting method is improved; an idea of a normalized cut framework system is adopted; the cut effect of the image is improved greatly; and the improvement on the division effect on the boundary of a sub-image is particularly remarkable.