摘要
图像压缩是提高图像存储效率以及实现高速高效传输的前提。根据神经网络的基本结构和算法,设计并搭建了基于卷积神经网络的CNNC(convolutional neural network compression,CNNC)图像压缩模型。该模型通过卷积层和池化层构成自编码器,反卷积层和卷积层构成自解码器,实现了图像编码压缩和解码重建的功能,并通过Set12数据集验证了CNNC图像压缩模型。实验结果表明,当压缩比较低时,JPEG压缩方法与CNNC压缩方法无显著差异;当压缩比较高时,CNNC压缩方法有明显的优势,在压缩比高达128时,CNNC压缩方法重建结果仍然很好。Set12数据集实验验证了CNNC压缩模型的有效性。
Image compression is the premise to improve the efficiency of image storage and realize high-speed and efficient transmission.According to the basic structure and algorithm of neural network,a convolution neural network compression(CNNC)model was designed and built in this paper.The self-encoder was composed of convolution layer and pooling layer,deconvolution layer and convolution layer constituted self-decoder.The function of image coding compression and decoding reconstruction was realized.The CNNC image compression model was validated by Set12 data set.The experimental results showed that when the compression ratio was low,there is no significant difference between JPEG compression method and CNNC compression method;when the compression ratio was high,CNNC compression method had obvious advantages,and when the compression ratio was up to 128,the reconstruction result of CNNC compression method was still very good.Set12 data set experiment verifies the validity of CNNC compression model.
作者
崔建良
李建飞
陈春晓
姜睿林
CUI Jianliang;LI Jianfei;CHEN Chunxiao;JIANG Ruilin(Department of Biomedical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)
出处
《生物医学工程研究》
2019年第4期415-419,共5页
Journal Of Biomedical Engineering Research
关键词
图像压缩
自编码器
卷积神经网络
深度学习
图像重建
Image compression
Self-encoder
Convolutional neural network
Deep learning
Image re-construction
