Study on a Construction of Transmission System for Super High Quality Images

通过本研究，揭示了数字博物馆应具备的功能。本学期是这项研究项目的最后一年，这项研究项目是科学研究助学金。研究结果总结如下：(1)揭示了可逆性、进步性和可伸缩性对于超高清图像应用系统的重要性。在这项研究中，我们提出了一种新的基于提升方案的自适应小波变换的可伸缩编码方法。此外，利用JPEG-2000的验证模型VM0对超高清图像进行了仿真实验。(2)为了保存数字博物馆中的三维物品，需要从多个方向拍摄超精细照片。此外，为了忠实地表现艺术作品，无损编码是必不可少的。为了实现这些功能，我们开发了一种新的编码方案，将从不同方向获取的物品的图像组进行存档，并允许自由阅读符合…更符合用户的需求。在我们的方案中，每一帧都可以被认为是旋转体和背景的组合。旋转器部分随着帧的改变而逐渐不同，并且因为该部分对应于原图的信息，所以在每一帧中该部分都需要无损编码。因此，我们首先提出了旋转体的提取和编码方法，然后利用帧内预测对每隔一帧的图像组进行编码，并同时考虑旋转角度，利用前一帧和后一帧的预测对其余帧进行编码。在使用帧间预测的帧中，来自前一帧的预测、来自后一帧的预测和帧内预测共同贡献了数据减少。因此，将一帧分割成块，每个块交替使用这三种预测，并提出了另一种编码方法。通过粘贴多视点图像集生成展开图像，并由该展开图像利用透视投影和亮度校正方法合成近似图像。然后对近似图像和原始图像的差值进行无损编码，并对两种方法进行了仿真和性能评估。较少

The functions required for digital museum was revealed through this study. This term is the final year of this research project, Grant-in aid for scientific research. Results of this study are summarized as follows ;(1) We revealed that the reversibility, the progressibility and the scalability are important for the super high definition image application systems. In this study, we proposed a new scalable coding method based on an adaptive wavelet transform with lifting scheme. Moreover, simulation experiments using the verification model VM0 of JPEG-2000 are performed for super high definition images. (2) To preserve a three-dimensional item in the digital museum, it is necessary to take super fine photographs from various directions. Moreover, in order to represent an artwork faithfully, lossless coding is indispensable. To realize these functions we developed a new coding scheme that an archive the image group of an item taken from various directions, and enables free perusal accord … More ing to the user's demand. In our proposal, each frame can be regarded as the composition of the rotator and the background. The rotator part differs little by little with frame changes, and because this part corresponds to the information of an artwork, lossless coding is required for this part in every frame. Therefore, we first develop the extracting and coding method for the rotator, and then intra-frame prediction is utilized to code the image group for every other frame, and the remaining frames are coded exploiting the prediction from previous and succeeding frame, taking account of the rotation angle simultaneously. In the frame using the inter frames prediction, prediction from previous frame, prediction from succeeding frame, and intra frame prediction together contribute the data reduction. Thus, a frame is segmented into blocks and each block uses the three kinds of prediction alternatively.Moreover, another coding method is developed. A developed image is made by pasting multi-view point image set and approximated images are composed from this developed image using perspective projection and luminance correction method. Then the difference from the approximated image and the original image are encoded in a lossless manner.Simulations of both methods are performed and performances are evaluated. Less