A method and system for processing image, a computer device and a readable storage medium. The method includes: images in an initial dataset are preprocessed to obtain a training dataset (S1); an image segmentation neural network is trained by the training dataset (S2); a last loss function layer of a trained image segmentation neural network is removed to obtain an inference network (S3); the training dataset is inputted into the inference network to obtain a plurality of logical vectors (S4); a check network is trained on a basis of the plurality of the logical vectors, the initial dataset, and a mask of each of the images in the initial dataset (S5); and inference is performed on an image to be processed by the inference network and a trained check network so as to obtain the mask of the image to be processed (S6).