Image enhancement method based on deep learning

Abstract

The invention discloses an image enhancement method based on deep learning, and the method comprises the steps: selecting an image shot by a professional photographer, enabling a professional retouching person to carry out the retouching of the image, constructing a neural network training data set, and dividing the neural network training data set into a training set Ttrain and a test set Ttest; adopting a U-Net neural network S(.) with global features; inputting an original image SINput subjected to data enhancement and priori illumination estimation I, and outputting enhanced Routput and Ioutput; randomly initializing related parameters such as weight parameters, learning rates and batch sizes in the neural network S(.); training the image enhancement neural network model by adopting an error back propagation algorithm, and calculating loss based on the weight map so as to obtain a depth image enhancement model. Compared with the prior art, the method has the advantages that the color is more natural, the method is more attractive, the contrast ratio is better, the difference between images modified by a professional image retoucher is smaller, the artifact phenomenon is avoided, the reasoning time is very short, and the method can be operated in real time at the ms level on equipment such as a mobile phone and the like.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image enhancement method based on deep learning. Background technique [0002] Photography is an art of light and shadow. A pleasing photo is an accurate grasp of the brightness of light and shadow and the shade of color. With the popularity of portable photography equipment such as mirrorless cameras and mobile phones, more and more people use photos to record their lives, start to enjoy the fun of photography, and are willing to share their photography works on social networks. Usually, people often complain about unnatural brightness, unsaturated colors, etc. in the photos they take. Therefore, people often spend a lot of time beautifying images before sharing their photographic works. Although there are a large number of interactive and semi-automatic image processing tools on the market, these tools still have a relatively large threshold for users, and the bea...

AI Technical Summary

Problems solved by technology

But these algorithms are only suitable for specific conditions, such as histogram equalization will not select data processing, automatic white balance is only suitable for uniform lighting conditions, etc.

And there is still a large gap between the processed image and people's expectations

In recent years, image processing algorithms based on deep learning have achieved great success in the field of image enhancement. However, these methods still have shortcomings such as slow processing speed and unsatisfactory results.

Images