New Standard for Still Image Coding - JPEG 2000

For a digital image, the amount of data may be up to hundreds of megabits. In order to efficiently and economically deliver images satisfying certain subjective requirements, certain measures must be taken to reduce the image data to be transmitted. Image (compression) coding is based on a large number of statistical analysis of digital images. It makes full use of the strong correlation of images, reduces the redundancy of image information and reduces the digital rate. In the past ten years, image coding technology has been rapidly developed and widely used, and several international standards for image coding have been developed, such as the ISO/IEC standard for still image coding JPEG (Joint Photographic Experts Group). CCITT (International Telephone and Telegraph Consultative Committee) video coding standard for videotelephone/conferencing H.261 and ISO/IEC coding standard for moving pictures MPEG series. These standard image coding algorithms incorporate a variety of traditional image coding methods with excellent performance. They are a summary of traditional coding techniques and represent the development level of image coding at the time. Among them, the coded JPEG standard for still images was released in 1990. The standard applies to black-and-white and color photos, color faxes and printed pictures, and can support higher image resolution and quantization accuracy, but is not suitable for binary images. It has become a universal standard widely used in image storage and communication. The JPEG standard defines two basic compression algorithms: 1. Discrete Cosine Transform (DCT)-based distortion-limited compression algorithm; 2. Spatial Linear Prediction (DPCM)-based distortionless compression algorithm. Four types of codec modes are allowed in JPEG: (1) Serial DCT mode; (2) Progressive DCT mode; (3) Distortion-free mode; (4) Hierarchical mode. Because of the shortcomings of the block DCT technology itself, the performance of this standard algorithm at higher compression ratios is not satisfactory, and it has not been able to adapt to the rapid development of video and multimedia technologies. Moreover, from the perspective of human visual characteristics, DCT-based transform coding can not provide the mechanism to make full use of human visual characteristics. In the late 1980s, research on the second-generation image coding technology that utilized human visual characteristics became increasingly active. The wavelet coding method not only has some advantages of the traditional coding methods, but also the multi-resolution transformation characteristics of the wavelet transform provide an effective mechanism for utilizing human visual characteristics, and is very promising in the field of high compression ratio coding. The extensive development of multimedia technologies, especially Internet technologies, poses new challenges for image compression and transmission technologies, including the requirements for progressive transmission of images. It is objectively required to propose and develop a new still image compression standard based on JPEG, JPEG2000. At the Sydney Conference on November 25, 1997, the recommendations made by the SAIC/University of Aricozona and USC/Rockwell were chosen as the starting point for a unified approach because of the best average decoding efficiency. Both image encodings are based on DWT (Discrete Wavelet Transform) rather than DCT. At a later meeting, the JPEG2000 will only be based on DWT image compression technology. JPEG2000 is not ready to replace the existing JPEG standard, and JPEG2000 will have higher coding efficiency and better performance. The significant difference between DWT and DCT is that the basis functions used are local. DWT is a good decomposition strategy because ordinary images usually do not have many periodic structures and can achieve higher coding efficiency. In addition, the unique structure of the wavelet coefficients allows it to generate spatial and quality scalable bitstreams, thereby enabling progressive transmission of images. In some bandwidth-constrained situations, such as a telephone line transmission network, downloading an image often requires a long waiting time. If this technique is used, a coarse-resolution image can be transmitted as quickly as possible. Gradually compensate for higher resolution or even distortion-free images as needed. This technology facilitates the browsing, querying, and transmission of databases. JPEG2000 is a kind of low-complexity and high-efficiency wavelet coding that can be used for image memory retrieval and transmission. More than 100 experts in more than a dozen countries are engaged in the development of the JPEG2000 standard. At the meeting held in Tokyo in March 2000, the first draft of the JPEG2000 was formally adopted. The formulation of the JPEG2000 Part 2 JPEG_LS lossless and quasi-lossless compression standard has also been completed. JPEG2000 has improved the original JPEG standard and added some new features. This is reflected in the following aspects: (1) Higher low bit rate performance: JPEG has good image quality at medium to high bit rates, but at low bit rates (for example, 0.25 bpp or less for high definition grayscale images) The quality has dropped significantly. JPEG2000 has been greatly improved in this regard, and it can meet applications such as network image transmission and remote sensing. (2) Compression of large-scale images: The image size allowed by the JPEG compression standard cannot be larger than 64K×64K. JPEG2000 basically does not limit the size of the image. JPEG2000 can also be used to compress infrared, multispectral, hyperspectral, SAR, and other remote sensing images. (3) Lossless and lossy compression: Previous standards cannot achieve lossless and lossy compression simultaneously in the same stream. JPEG2000 can get lossy or lossless images during progressive decoding. This is very meaningful for some applications. For example, medical claims image, we usually want to losslessly store it, and in some cases show lossy; in the network environment, the user can choose different display methods for different display terminals. (4) Progressive transmission with pixel accuracy and resolution: The so-called progressive transmission refers to reconstructing images with increasing pixel accuracy or spatial resolution. It can be used in WWW, image file systems, printers, and the like. This is one of the most important features of JPEG2000. (5) Applicable to fixed-rate channels, fixed stream size, limited hardware working memory applications: JPEG2000 allows code streams to pass through a limited bandwidth channel at a fixed rate, for example, real-time transmission of remote sensing images, moving image encoding Wait. JPEG2000 can control the stream length regardless of the actual size of the image, which can meet the needs of some fixed-memory hardware, such as scanners, printers, and so on. (6) Random stream reading and processing: Usually an image contains some relatively important information compared to other parts. The user can define some regions of interest in the diagram so that it can be read randomly or when it is decoded with little distortion, and operations such as rotation, translation, filtering, and feature extraction can also be performed. (7) Robustness: In the past, bit error occurred in the standard stream transmission, and the quality of the picture dropped sharply or could not be recovered. The JPEG2000, due to its special code stream structure, can still recover images well in case of bit errors. (8) Compression of continuous tone and binary images: Due to the poor compression performance of binary (text) images, JPEG is rarely used for the compression of composite files. JPEG2000 can compress continuous tone and binary images in the same encoding system. For example, JPEG2000 can compress composite images containing text and other information, medical diagnostic images containing annotations, and graphics and computer-generated binary images. Due to the above characteristics, the JPEG2000 compression standard can be widely used in many occasions such as facsimile machines, printers, scanners, digital photography, remote sensing, remote medical diagnosis, and Web database access. (Author: Zhengzhou University of Information Engineering, Graduate School of Surveying and Mapping Training Center) .Editor Li Yonglin.