摘要
波域图像的子带区域划分混乱,导致图像编码序列偏移量过高,图像压缩编码精度较低等问题。基于此,提出基于CNN的高动态范围图像压缩稀疏编码方法。采集高动态范围图像压缩稀疏编码样本,将波域图像的子带区域划分成多个坐标集合,利用图像压缩数据处理稀疏编码样本字符集,确定单点子带与系数幅值的差异,在此基础上,利用编码数据合成频带稀疏矩阵,并借助卷积神经网络训练编码数据,完成高动态范围图像压缩系数编码。实验结果表明:文中方法的图像编码序列偏移量维持在0.2E上下,且有效提升了高动态范围图像压缩编码的精度。
The sub-band division of wave region image is chaotic,which leads to the problems of high offset of image coding sequence and low accuracy of image compression coding.Therefore,a high dynamic range image compression sparse coding method based on CNN is proposed.Firstly,the high dynamic range image compression sparse coding samples are collected,the sub-band region of the wave region image is divided into multiple coordinate sets,and the sparse coding sample character set is processed by the image compression data.Then,the difference between single sub-band and coefficient amplitude is determined.On this basis,the frequency bandwidth sparse matrix is synthesized by the coded data,and the coded data is trained by convolutional neural network to complete the high dynamic range image compression coefficient coding.The experiment results show that the image coding sequence offset of this method is maintained at about 0.2 E,and the accuracy of high dynamic range image compression coding is effectively improved.
作者
张继东
曹靖城
李云鹤
ZHANG Ji-dong;CAO Jing-cheng;LI Yun-he(Tianyi Smart Home Technology Co.,Ltd.,Nanjing 210001,China)
出处
《信息技术》
2022年第12期123-129,共7页
Information Technology
关键词
高动态范围
图像压缩
稀疏度
编码
high dynamic range
image compression
sparsity
coding
