Study on a Basic Concept of Highly Efficient Coding Model for Super High Definition Images

随着硬件性能的提高和多媒体内容的增加，高分辨率和像素精度的图像得到了广泛的应用。在具有原始分辨率的普通尺寸显示器上，很难完全显示大尺寸图像，如超高清(SHD)图像。在这种情况下，首先显示缩小图像以让用户选择他希望看到细节的感兴趣区域。然后从码流中提取所选区域的图像数据。因此，人们非常希望引入一些部分可译码的码，利用这些码可以独立地译码码流的任何部分。为了实现这种可伸缩性，我们将图像分割或分解成大小合适的块或瓦片，并分别使用DPCM方法或正交变换对每个块进行编码。然而，如果将变长码用于熵编码，则很难检测到码序列中每个块的开始位置。在这项研究中，我们提出了一种新的无损编码方法，能够显示图像的感兴趣区域，而不需要从编码序列的起始点进行解码。我们还提出了一种可部分译码的可变长码--修正Golomb-莱斯码。

Images of high resolution and pixel accuracy are coming into wide use with the improvement of hardware performance and the increase in multimedia contents. It is difficult to represent a large size image entirely, such as a super high definition (SHD) image, on an ordinary size monitor with the original resolution. In this case, a reduced image is displayed at first to let user select a region of interest where he hopes to see the detail. Then image data of the selected area are extracted from code stream. Hence, it is highly desired to introduce some partially decodable codes with which any parts of the code stream can be decoded independently. In order to realize such scalability, we divide or decompose an image into blocks or tiles of proper size, and encode each block respectively by DPCM method or orthogonal transform. However, it is difficult to detect the start position of each block in code sequence if variable length code is used in entropy coding. In this study, we propose a new loss less coding method enable to display the region of interest of the image without decoding from the start point of the code sequence. We also propose apartially decodable variable length code named modified Golomb-Rice code.