您当前访问的的浏览器版本过低,为了给您带来更好的体验,建议您升级至Edge浏览器或者推荐使用Google浏览器
取消

视频质量增强模型加速算法

作者:杨文哲,徐迈,白琳 阅读量:1191

视频质量增强模型加速算法

杨文哲,徐迈,白琳
(北京航空航天大学,中国 北京 100191)

摘要:提出了一种应用于视频质量增强算法的动态结构性剪裁算法Maskcut,它可以有效提高基于深度学习的视频质量增强算法的运行速度。Maskcut是一种通用的剪裁思路,支持绝大多数的基于卷积神经网络(CNN)深度学习网络模型的剪裁加速。基于原模型中已经训练好的参数数据,Maskcut使用一种针对剪裁加速的二次训练策略来进一步微调参数,从而在保证模型有效性损失不大的同时,追求模型运行时间的缩短。以一种先进的视频质量增强算法MFQE 2.0为目标,Maskcut剪裁后可以快速达到峰值信噪比(PSNR)指标损失低于1%、时间缩短10%以上的加速指标。 
关键词:模型加速;图像质量增强;结构性剪裁 


Video Quality Enhancement Model Acceleration Algorithm

YANG Wenzhe, XU Mai, BAI Lin
(Beihang University, Beijing 100191, China)

Abstract: Maskcut, a dynamic structural clipping algorithm for video quality enhancement algorithm is proposed, which can effectively improve the speed of video quality enhancement algorithm based on depth learning. Maskcut is a general tailoring idea that supports most of the tailoring acceleration based on convolutional neural networks (CNN) deep learning network models. Based on the trained parameter data in the original model, the secondary training for tailoring acceleration is carried out to further fine-tune the parameters. With an advanced video quality enhancement algorithm multi-frame quality enhancement (MFQE) 2.0 as the goal, the Maskcut can quickly reach the acceleration index with peak signal-to-noise ratio (PSNR) index loss of less than 1% and time reduction of more than 10% after trimming. 
Keywords: model acceleration; image quality enhancement; structural tailoring

在线PDF浏览:PDF