一种高精度的自适应码率控制图像压缩算法  

An Image Compression Algorithm With High-Precision Adaptive Rate Control

作　　者：宋蓓蓓[1] 何帆 马穗娜 孙文方 SONG Beibei;HE Fan;MA Suina;SUN Wenfang(School of Information Engineering,Chang’an University,Xi’an 710064,China;School of Aerospace Science and Technology,Xidian University,Xi’an 710126,China)

机构地区：[1]长安大学信息工程学院,西安710064 [2]西安电子科技大学空间科学与技术学院,西安710126

出　　处：《西安交通大学学报》2022年第2期198-206,共9页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(62001059);陕西省重点研发计划资助项目(2021GY-019)。

摘　　要：针对非完全嵌入编码的图像压缩算法难以实现精确的码率控制问题,提出了一种结合码率预测和压缩后率失真优化的高精度自适应码率控制的图像压缩(HTJ2K-RPRD)算法。首先,根据二维离散小波变换高频子带数据的高斯分布特性,计算量化前高频子带数据的信息熵得到最小平均比特数,用最小平均比特数来估计量化编码后高频子带数据的码率预测值;接着,采用与JPEG2000标准压缩质量相当但复杂度比其低的HTJ2K标准进行图像压缩,根据HTJ2K标准中清理通道的编码特点,粗略计算二维离散小波变换低频子带数据在码流中嵌入的平均比特数,用嵌入的平均比特数来估计编码后低频子带数据的码率预测值;然后,统计所有编码后子带数据的码率预测值来估计基础截断位平面;最后,从基础截断位平面编码产生多个编码通道,对产生的编码通道进行率失真优化以实现自适应码率控制的图像压缩。实验结果表明:HTJ2K-RPRD算法在自适应码率控制的情况下可实现带宽受限压缩系统的定码率压缩,且复杂度更低;在不同位深和像素大小的图像上的压缩码率控制精度可达99.997%。An image compression(HTJ2 K-RPRD) algorithm with high-precision adaptive rate control is proposed to solve the problem that the image compression algorithm with incomplete embedded coding is difficult to achieve accurate rate control, and the algorithm combines bit rate prediction and post compression rate distortion optimization. Firstly, according to the Gaussian distribution characteristics of the high-frequency subband data of two dimentional discrete wavelet transform, the information entropy of the wavelet coefficients before quantization is calculated to obtain the minimum average bit number which is used to estimate the predicted bitrates of the high-frequency wavelet coefficients after quantization and encoding. Secondly, the HTJ2 K standard, which has the same compression quality as the JPEG2000 standard but a lower complexity than it, is used for image compression. According to the coding characteristics of the cleanup passes in the HTJ2 K standard, the average bits embedded in the code stream of the low-frequency wavelet coefficients are roughly calculated and the predicted rates of the encoded low-frequency wavelet coefficients are estimated by the average embedded bits. Then the predicted bitrates of all encoded wavelet coefficients are counted to estimate the basic truncated bit plane. Finally, multiple coding channels are generated from the base-truncated bit-plane coding, and the generated coding channels are optimized for rate-distortion to achieve adaptive rate control image compression. Experimental results show that the HTJ2 K-RPRD algorithm can achieve fixed rate compression for bandwidth-constrained compression systems with lower complexity under the condition of adaptive bit rate control;the compression bit rate control accuracy can reach 99.997% on images with different bit depths and pixel sizes.

关 键 词：图像压缩 码率控制 编码 率失真优化 JPEG2000标准 HTJ2K标准 

分 类 号：TN911.7[电子电信—通信与信息系统]