利用多尺度条件熵和记忆性度量估计图像无损压缩极限

Estimation of lossless image compression bounds based on multi-scale conditional entropy and memory measurement

为估计利用各类成像传感器自然获取图像的无损压缩极限 ,提出一个利用多尺度条件熵和记忆性度量的实用方法。传统方法进行高阶概率估计时由于上下文稀释问题而需要大量的训练数据 ,该方法借助多尺度的策略有效地缓解了这一个问题 ,从而可以利用有限的观测数据进行高阶条件熵分析 ,进而对无损压缩极限作出更为准确的估计。已进行的实验结果表明 ,该文提出的估计方法可以较好地适用于扫描仪、遥感、医学等各类成像传感器 ,为相关应用提供了可靠参考。

Lossless compression bounds of images naturally acquired by various imaging sensors are estimated using a practical method based on multi scale conditional entropy and memory measurement. The multi scale solution can efficiently resolve the context dilution problem usually occurs in high order statistical estimates. Thus, high order conditional entropy analysis can be carried out with an ordinary image as opposed to dealing with an enormous, practically unmanageable training set. This leads to more precise estimates of lossless compression bounds for real images. The feasibility and reliability of the proposed method have been verified on a group of images acquired by scanners, medical imaging sensors and remote sensors.