METHOD AND DEVICE FOR AUTOMATICALLY CROPPING PICTURE BASED ON REINFORCEMENT LEARNING

摘要

A method and device for automatically cropping a picture based on reinforcement learning, which belong to the field of image processing. The method comprises the steps of: using a reinforcement learning model to perform feature extraction on a current cropping window to obtain a local feature, splicing the local feature with a global feature of a picture to be cropped to obtain a new feature vector, and using the new feature vector as current observation information (S1); using an LSTM unit in the reinforcement learning model to record historical observation information, and combining the historical observation information with the current observation information to serve as a current status representation (S2); and according to a cropping strategy and the current status representation, performing a cropping action on the picture to be cropped in a serialized manner to obtain a cropping result (S3), wherein the reinforcement learning model is a model constructed based on a convolutional neural network. By means of the method, an accurate picture cropping result can be quickly acquired.