@misc{oai:niigata-u.repo.nii.ac.jp:00004738, author = {Ishida, Takuma}, month = {Mar}, note = {The goal of this study is to develop non-separable multi-rate systems integreted with invertible video format coversion in order to offer a spatio-temporal resoluton control function as a new option for scalable video coding., Chapter 1 describes a background of this study. As a comprehensive consequence of the explosive growth of Internet, great advaces in hardware technologes and software developments, a lot of new multimedia applications are emerging rapidly. Although the stroge capacity of digital devices and the bandwidth of networks are increasing constantly and rapidly, video compression still plays an essential role in these applications. In the applications of video delivery services such as video streaming over Internet and mobile communication, it is important to offer dynamic adapability to a diversity of different networks, terminals and video formats with a seamless bitstream that conveys the video content. In this situation, a scalable function provides adaptive contorl capability of images quality (i.e. SNR), temporal rate, and picture size (resolution). Scalable video coding is an attractive way to universal and seamless access to video contents with dynamic controllability that meets diffferent requirements in different networks, terminals, formats and displays including a TV and a PC. Given these backgrounds, one of the hard problems in scalable video coding is a format conversion between interlaced and progressive scanning. The motivation of this study is to deal with a coding system using a new class of deinterlacing, which is referred to as invertible deinterlacing, for motion picture coding. Deinterlacing is a kind of sampling rate conversion of video signals. Usually it is defined as the sampling rate up-conversion of an interlaced scan signal to its progressive scan counterpart by interpolating the missing lines in every field. The deinterlaced signal will have twice the vertical sampling rate to the original. The invertible deinterlacing technique is defined in a new form different from the conventional deinterlacing techniques. In applications of scalable video coding, twice the video infomation data to the original is undersirable before encoding. Given an interlaced scanning video signal, the invertible conversion produces its corresponding progressive scanning video signal, where both of these two signals have the same sampling density, while maintaining the perfect reconstruction property. The objective of this study is to develop coding systems equipped with new spatio-temporal scalability. In the following chapters, several multirate systems with the invertible conversion are discussed so as to find a better solution., Chapter 2 presents a lossy implementation scheme of Motion-JPEG2000 (MJP2) integrated with invertible deinterlacing.. The invertible deinterlacing technique is developed in order to suppress comb-tooth artifacts which are caused by field interleaving for interlaced scanning videos and affect the reconstructed picture quality in scalable frame-based codecs such as MJP2. The technique is characterized by two features. The sampling density is preserved and image quality is perfectly recovered by an inverse process. When no codec is placed between the deinterlacer and inverse process, the original video is perfectly reconstructed. Otherwose, it is almost completely recoverd. An application scenario of this invertible deinterlacer is also presented for enhancing the sophisticated SNR scalability in the frame-based MJP2 coding. The proposed system suppresses the comb-tooth artifacts at low bitrates, while enabling the quality recovery through its inverse process at high bitrates within a standard bitstream format. The main purpose of this chapter is to present a system that yields high quality recovery for an MJP2 codec in lossy compression. We demonstrate that our invertible deinterlacer can be embedded into the discrete wavelet transform (DWT) employed in MJP2. AS a results, the energy gain factor to control rate-distortion (R-D) characteristics is compensated for an optimal compression. Finally, the performance evaluation of image recovery by the proposed method is discussed. Several test sequences and filter characteristics are investigated through the simulation results., Capter 3 describes a lossless compression technique of MJP2 integrated with the invertible deinterlacers. The proposed system suppresses comb-tooth artifacts in frame pictures generated from interlaced pictures, and is applicable to both of frame and field based-displaying applications with a single code-stream. It is easily imagined that those frame picture are frequently edited and recompressed regardless of being used in professional or consumer applications. Usually, its is required to preserve all accurate information during such repetitive processes. The reversible, i.e. lossless conversion to reconstruct interlaced images from deinterlaced code-streams is therefore of significant interest. Lossless compression is desirable in these applications where images are likely to be edited and recompressed reversible times, whereas the accumulation of errors by multiple lossy compression can grow to an intolerable level. This chapter describes an exact lossless implementation technique. The proposed implementation is realized by integrating the deinterlacer into MJP2 codecs and modifying the header infomation for appropriate standard decoding. Some simulation results show that the comb-tooth suppression capability is kept at low bitrates with standard MJP2 decoders., Chapter 4 discusses motion compensated three-dimensional (3-D) filter banks providing a new tool of a spatio-temporal resolution control for scalable video coding. There are recent developments in scalable video coding, and most of them are yet based on 3-D wavelet transform with motion compensation. To achieve the function of frame-rate and spatial resolusion scalabilities, motion compensated temporal filtering (MCTF) through a lifting wavelet transform currently attracts many reserchers as an effective temporal decomposition tool. To provide the video format control between the progressive and interlaced scanning, two different structures with a spatio-temporal split process are presented as a unique spatio-temporal scalability. Unlike other filter banks, the proposed systems are construted in a unique way to multi-dimentional (VT) quincunx and face-centered-orthombic (FCO) format. At first among them, perfect reconstruction (PR) deinterlacer banks are suggested. The propsed system separates a progressive video into a pair of different progressive videos of a half frame rate. It is novel from the viewpoint of filter banks where interlaced videos are given as intermediate data during analysis and synthesis process. Unlike the coventionals filter banks, our systems are constructed in a way unique to multidimensional systems by using invertible deinterlacers which we have proposed. This technique offers a functionality by which a reconstruted base layer itself provides both of half-rate progressive and interlaced videos, and an additional enhancement layer, which has a prediction error in high-pass subbands, improves the spatiotemporal resolution. Secondly, a novel non-separable motion compested spatio-temporal filter (MCSTF) is proposed. This technique can provide the video format control between progressive and interlaced scanning for spatial-temporal scalable video coding as well as deinterlacer banks. This technique corresponds motion compensated temporal filtering, which has an important role as a component of 3D-DWT for exiting scalable video coding techniques. As a new tool, we proposed a motion compensaton technique for non-separable sampling lattices in terms of spatial-temporal decompotion, The proposed method has a different structure with motion compensated field prediction from the deinterlacer banks. In some experimental results, we show significance of these proposed techniques. Finally in Chapter 5, conclusion of each chapter is drawn after an entire summary of this study., 新大院博(工)甲第225号, 新大院博(工)甲第225号}, title = {Scalable Video Coding with Invertible Sampling Lattice Conversion}, year = {2006} }