A Study on Data Compression Using Image Sets and Ambiguous Entropy

二维数字单色图像中的信息可以看作是从像素域到对比度强度区域的映射函数。它通常用矩阵表达式表示。可以认为，JPEG等图像编码通过简洁地表示近似类型，实现了信息压缩。本研究提出的图像编码方法对原始图像中每个像素的强度范围内的数据进行处理，该强度范围有上界和下界。它具有适合对象编码、能够使用集合运算、能够定量、系统地处理质量层次等特点。利用模糊信息的概念，在更新处理过程中能够基于较优准则再现图像。在本报告中，物体的类别采用了表示光滑区域的光滑图，识别物体边界的边缘图，以及不依赖于图片种类，可以应用于任意域的一维扫描图。对自然图像和单色或彩色插图图像进行了图像压缩和重构，详细分析了该方法的优点和缺点。在彩色图像中，我们使用相互独立的YUV表颜色系统对平原进行分离。对每幅图像进行编码和解码，合成图像。通过各种实验，可以证实上述特征有效地发挥了作用。但是在自然图像中存在着压缩性能不如JPEG的缺点。在今后的改进中，有必要增加新的对象类，并设计彩色图像的表达模式。

Information in two-dimension digital monochromatic image can be considered to be a mapping function from the domain of pixels into the region of the intensity at contrast. It usually shows by the matrix expression. It can be thought that the image encoding such as JPEG has achieved the information compression by concisely expressing the approximation type. The image encoding method proposed by this research handles data within the range of the intensity in each pixel of the original image which has the upper bound and the lower bound. It has features such as suiting the object encoding, being able to use the set operations, being able to treat the level of the quality quantitatively and systematically. by using the concept of ambiguous information, and being able to reproduce the image based on a preferable criterion during the renewal processing.In this report, the classes of the objects adopt the smooth figure where smooth area is expressed, the edge figure where boundary of an object is identified, and the one-dimension scanning figure which is not depending on the kind of pictures and can be applied to arbitrary domains. The image compression and the reconstruction were tried to various images like a natural image and an illustration image with monochromatic or e color, and the advantage and the fault of the proposed method were investigated in detail. In the color image, we separate the plains by using the YUV table color system which are mutually independent. This encoding and decoding process are performed to each, and the constructed picture is compounded. It is able to be confirmed that the above-mentioned features functioned effectively as the result of various experiments. However, there is a fault that the compressibility is inferior compared with JPEG in natural images. As future improvements it is necessary to add the new object classes and to devise the mode of the color image expression.