摘要
基于第二代Bandelets变换提出了一种静止图像压缩算法．先引入数字线表示离散域中的直线，再通过目标码率来计算相应的近似量化值，对建立的率失真函数取最小值得到最佳的Bandelets变换方向，然后根据子带相关性以建立四叉树并通过编码扫描次序来存储系数，最后将JPEG2000的比特位平面和上下文编码应用到量化后的Bandelets系数，从而实现了图像压缩．该算法降低了方向寻优的计算复杂度，减少了表征方向信息所需比特数和辅助的方向信息总量，同时提高了压缩图像的质量．实验结果表明，与JPEG2000相比，该算法的峰值信噪比可提高0．3-1．3dB．
Based on the second generation Bandelets transform, a novel still image compression algorithm is presented. Firstly, digital line is introduced to describe the straight line in discrete domain. Secondly, a proximate quantized value is computed according to the object encodingrate. The direction of the best Bandelets transform is obtained by minimizing the constructed rate-distortion function. Then the quadtrees are built according to the sub-band correlation. And the coefficients are stored by using the coding scan order. Finally, bit plane and contextual coding procedure of JPEG2000 scheme are applied to quantized Bandelets coefficients to realize image compression. The proposed algorithm reduces computation complexity of optimal direction search and improves the quality of image compression. Experiments show that, comparing with JPEG2000 scheme, peak signal to noise ratio gets an improvement of 0. 3 dB to 1.3 dB.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2006年第8期950-954,共5页
Journal of Xi'an Jiaotong University
基金
"十五"国防预研基金资助项目(41321090210)
作者简介
孙文方（1979-），男，博士生；
赵亦工（联系人），男，教授，博士生导师．
