摘要
图像修复是数字图像处理中的一项经典任务,随着深度学习技术步入人们的视野,计算机修补残缺图像的能力得到大幅提升。本文介绍了如何利用最小二乘损失,改进原始的生成式对抗网络,预测残缺图像中的丢失信息并快速修补残缺图像,克服传统图像复原方法中存在的问题。
Image inpainting is a classic issue in digital image processing.Computers’ability of image restoration has been greatly improved as deep learning technology walk into people’s vision.This article describes how to overcome the problems in traditional inpainting methods by using latest square loss to improve the original generative adversarial network with deep convolution and predict the missing part of a residual image to do image inpainting efficiently.
出处
《科技视界》
2020年第22期1-6,共6页
Science & Technology Vision
