Sensuous Assesment Function of Images Based on Local Entropy Flow

本研究的主要目的是利用局部熵流的概念定义图像的感官评价函数，并提出一种新的图像感官编码方案。首先，引入局部熵流的概念，并将其定量定义应用于一些人工图像。用三维表示法绘制了熵流图，证明了上述定义的正确性。其次，我们定义了一个新的概念，熵滤波器的控制参数的传统的低通滤波器的图像根据每个值的局部熵流。然后将熵滤波器应用于标准图像集SIDBA，并利用其结果详细研究了熵滤波器的滤波性能。最重要的结果是，在滤波图像的信息量大大减少，但感官退化是很少的，因此，如果熵滤波器作为矢量量化的预处理器，一个新的编码方案，具有很低的速率，但很少的感官退化将成为可能。最后，基于局部熵流定义了图像的感官评价函数，并对标准图像集“SIDBA”进行了评价。结果表明，该评价函数虽然只是相对值，但很符合视觉特性。

The main object of this research is to define the sensuous assesment function of images using a new concept of local entropy flow, and to develop a new sensuous coding scheme of images. First, a new concept of local entropy flow is introduced, and its quantitative definition is applied to some artificial images. The obtained entropy flow is plotted by three dimensional representation from which the validity of the above definition is certified. Next, we define a new concept of entropy filter which controls the parameter of the conventional low pass filter for images according to each value of the local entropy flow. Then the entropy filter is applied to the set of standard images "SIDBA", and using the results the filtering property is studied in details. The most important result is that the amount of information in the filtered images is greatly decreased, but sensuous degradation is very few, and hence, if the entropy filter is applied as a preprocessor of vector quantization, a new coding scheme with very low rate but few sensuous degradation will become possible. Finally, the sensuous assesment function of images is defined being based on the local entropy flow, and is evaluated for the set of standared images "SIDBA". The results show that the assesment funtion fits very well in the property of visual sensations though they are only relative values.