Reinforcement learning based automatic picture cutting method and device

Abstract

The invention provides a reinforcement learning based automatic picture cutting method and device, relates to the field of image processing, and aims at obtaining an accurate picture cutting result rapidly. The method comprises the steps that a reinforcement learning model is used to carry out feature extraction on a present cutting window to obtain a local feature, the local feature is spliced toa global feature of the picture to be cut to obtain a new characteristic vector, and the new characteristic vector serves as present observation information; historical observation information obtained via the reinforcement learning model and the present observation information are combined and serve as present state representation; and according to a cutting strategy and the present state representation, a cutting motion is carried out on the picture to be cut sequentially, and a cutting result is obtained. The reinforcement learning model is a model which is established on the basis of a convolutional neural network. The method and device can be used to obtain the accurate picture cutting result rapidly.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method and device for automatically cropping pictures based on reinforcement learning. Background technique [0002] With the rapid development of the image processing field, image automatic cropping, as an important part of the image processing field, has also been greatly improved. Image automatic cropping requires the computer to automatically pick out well-composed regions from the input image, which have higher aesthetic quality than the original image. [0003] The traditional method is based on the sliding window method, which mainly extracts candidate regions through sliding windows, and then extracts features from each candidate region and scores them, and the region with the highest score is used as the final result. However, this method will generate a large number of candidate windows, and the shape and size of each candidate window are relatively fixed. The croppin...

AI Technical Summary

Problems solved by technology

However, this method will generate a large number of candidate windows, and the shape and size of each candidate window are relatively fixed. The cropping window with a fixed shape and size cannot cover all situations, so not only the cropping effect is not good, but also it will consume a lot of calculations. resources and longer

Images