摘要
针对JPEG2000标准建议的码率控制算法的计算复杂度高且占用存储空间大的问题，提出了一种基于线性预测的动态阈值截断的码率控制算法．通过分析小波变换特点及图像边缘和纹理特性，根据图像的压缩倍数线性预测码流截断点的率失真斜率阈值，在编码过程中动态调整此阈值以满足码率的要求．该算法实现了码流的实时截断，降低了计算量和存储空间，提高了编码器的编码效率，保持与原算法相近的图像质量．该算法与JPEG2000码率控制算法相比，峰值信噪比（PSNR）只下降了0．1～0．3dB，在图像压缩倍数大于8的情况下，计算复杂度下降了41．6％以上，存储空间节省了43．2％以上．
A dynamic threshold rate control algorithm based on linear prediction was proposed to resolve the computation complexity and large memory usage in JPEG2000 rate control algorithm. By analyzing the characteristic of wavelet transform and the image＇s edge and texture features, the truncation point thresholds of rate-distortion slope were predicted linearly based on the image data compression ratio, which were modified dynamically to meet the requirements of the bit-rate during the encoding procedure. The algo- rithm real-time truncates the code-stream, adequately reduces the coding complexity and memory size, meanwhile keeps nearly the same image visual quality and speeds up the encoder＇s performance. Compared with the conventional algorithm in JPEG2000 standard, the proposed algorithm has approximate 0.1 0.3 dB reduction in peak signal-to-noise ratio （PSNR）, and above 41.6% computation and 43.2% memory usage can be decreased in the case of compression factor greater than 8.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2008年第8期1335-1339,1344,共6页
Journal of Zhejiang University：Engineering Science
基金
浙江省重大科技计划资助项目(021101559)
霍英东教育基金资助项目(94031)
