Advanced Scalable Coding And Motion Estimation Algorithms For Error Resilient Video Compression

The demand for media rich applications over the Internet as well as wireless networks has rapidly increased. Although these media communication and entertainment services are useful, however they are also the most bandwidth-consuming modes of communication. To meet the requirements of different users in different applications, not only high broadband access, but also an enhanced media coding methods are necessary to make the transmission more efficiently. Several video coding standards, such as the ISO/IEC MPEG series and the ITU-T video coding standards have been successfully developed and can significantly reduce the data rate. Most of these video compression methods use a block-based Discrete Cosine Transform (DCT) with motion-compensation to remove spatial and temporal redundancy.Now we have the demand for the transmission of video in real time with good quality. To get such service increasing the bandwidth is the easiest solution. However it is not cost effective. Reducing the amount of data to be transmitted means increasing the capacity of the communication channel, and similarly reducing a data to half of its size is the same as doubling the capacity of the storage medium. This fact emphasizes that reducing the amount of data to be transmitted is as an option for us. This leads to nowadays the state-of-the-art video/image compression standards.The most crucial part of video compression is motion estimation. Motion estimation is the process which generates the motion vectors that determine how motion compensated prediction frame is created from the previous frame. But its computational complexity poses great challenges for real time implementations. Among different motion estimation algorithms block matching motion estimation is the mostly applied method. To minimize the search time on block matching, a simplified and efficient algorithm is very crucial for fast motion estimation process. In this thesis the performance evaluation of different block matching algorithms (Exhaustive Search or Full Search (FS), Three Step Search (TSS), New Three Step Search (NTSS), Four Step Search (4SS), Diamond Search (DS), Adaptive Rood Pattern Search (ARPS) are presented and compared based on the metrics of Peak Signal Noise Ratio (PSNR), Mean Absolute Difference (MAD) and Search Points per Macro block. Moreover the ARPS algorithm which performs the best among the above mentioned algorithms is introduced and discussed. In addition to this, based on ARPS, a new dynamically adaptive Rood Pattern search Algorithm which utilizes the Spatial correlations of neighboring blocks (hence denoted it by ARPS_S to differentiate from ARPS) is proposed. The ARPS_S algorithm out performs the ARPS.Due to successful development of the video compression techniques, the huge amount of redundant information contained in the raw videos can be effectively removed. This makes it possible to transmit videos over limited-bandwidth channels. However, the highly compressed videos are fragile to noisy channels. Once compressed video bit streams encounter any kinds of errors in transmission, the quality of the reconstructed videos at the decoder side degrades seriously.In order to enhance the quality of the decoded videos at the decoder side, two common techniques are often adopted:one is to generate robust compressed video bit streams at the encoder side, which is known as the "error resilience" video coding; the other is to conceal errors in the reconstructed videos at the decoder side, which is known as the "error concealment" method. The error resilient technique from the encoder side includes layer coding with transport prioritization, multiple descriptions coding and robust entropy coding. The post-processing error concealment method is a technique at the decoder side.Robust video coding plays an important role in limiting the error propagation and improving visual quality in case of errors. It addresses the issue of error concealment by designing proper structures and maintaining acceptable redundancy while minimizing the complexity.In this thesis we focus on the error resilient technique using the scalable coding strategy. Scalable video coding is a coding technique which encodes a video sequence into a set of bit streams that can support various levels of qualities at the decoder side. Scalability provides robustness to video in which certain types of loss are acceptable and may not terminate decoding or severely affect visual quality. Layered scalability and parallel scalability (Multiple Description Coding (MDC)) are the two types of methods that benefit the video transmission with and without unequal protection. In layer coding (LC) case, the base layer is more important than the enhancement layers. In MDC the entire bit streams (descriptions are equally important). One part of the research work of this thesis is to study how to generate the bit steams in the LC case and the MDC case respectively. We also make performance comparison between these two techniques. Simulation results show that the MDC case outperforms the LC case.