Font Size: a A A

Unequal Error Protection Research Based On H.264 Code Stream

Posted on:2008-08-05Degree:MasterType:Thesis
Country:ChinaCandidate:R G ZhouFull Text:PDF
GTID:2178360212496917Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
H.264 is a new video coding standard. It's developed by united video group JVT which consists of Video Coding Experts Group(VCEG) of ITU-T and Movable Image Coding Experts Group(MPEG)of ISO/IEC. It uses many new coding techniques and can improve coding efficiency obviously.H.264 inherits mature error code resistance technologies from various kinds of previous coding standards. At the same time, it improves error code resistance technologies which include intraframe coding, data partition and so on, and puts forward new error code resistance technologies which include parameter set, flexible macro block sequence, redundant piece and so on. Furthermore, it also adopts the structure and syntax called"friendly network", and divides the whole framework into VCL layer and NAL layer, which also enhances the adaptive performance of various kinds of channels and is helpful to control error code and packet loss.In recent years, with the fast development of communication technologies and the continuous increase of individual user's demands for new services in communication, multimedia communication technologies, especially video communication technologies develop quickly. However, due to the huge video data information, compressed encoding must be executed to reduce the information quantity of channel transmission before transmitting it. The video information after compressed encoding is easily influenced by the channel condition and other factors, so it may produce error code easily during transmission. To guarantee the quality of video image reconstructed by the decoder, it needs some error control technique (ECT) to enhance the performance of resisting error code during transmission after encoding the video information.As a kind of strategy guaranteeing the reliable transmission of video information in channels, unequal error protection gets extensive and deep research. This strategy mainly studies how to classify the information produced by the source encoding terminal and makes the important data bits own stronger error recovery ability by adopting corresponding unequal protection to achieve the goal of making the entire decoding best.While H.264 encoding, encoding mode is designed to partition mode, which means that each data partition's encoding information is stored in different data partition according to the importance of them. The important information such as the header message of slice and macroblocks, moving vector information and macro block information is stored together, while the unimportant information like the encoded mode information of data block and the transformation coefficients of macro block in intraframe coding and interframe coding is stored in addition. Therefore, it can use unequal error protection based on the importance of the information in each several part. The important information gets error code protection with high priority while the unimportant information gets error code protection with low priority.This paper puts forward an unequal error protection plan based on XOR A. The purpose of the plan is to get good balance between error protection and redundancy consumption. That's to say, error protection isn't purely at the cost of redundancy. At the same time as error protection for the code stream produced by the H.264 encoder, it also need consider coding efficiency to get good balance between code rate and quality.The core idea is implementing unequal error protection with the method, which is combining DPM with increasing redundant XOR A packets for type A data packets, when channel packet loss rate is relative high. All RTP packets of partition A which belong to the code stream outputted from encoder are stored temporarily in reserved buffers. Then these RTP packets are encapsuled. Bit-oriented XOR is done for all RTP packets of partition A from the same packet. After that the bit stream through XOR is packeted and then the redundant XOR A packets are produced which are used for recovery when other A packets of the same packet are lost. We can identify easily the redundant XOR A packets in the decoding terminal through designing their structure and operating them. Furthermore, we can dynamically adjust the length of packet according to the packet loss rate. When the packet loss rate increases, the length of the packet should be decreased and the protection degree of partition A should be strengthened. The decrease of the packet loss rate can increase the length of partition to reduce the redundancy. Under the condition of the low packet loss rate, using NoDPM in the encoding terminal can save bit outgoings.The experimental result indicates that, under the condition of the same packet loss (for example, 3% packet loss rate), PSNR Y based on XOR A plan is 1.8db higher than NoUEP(which doesn't adopt unequal error protection plan), and 0.5db higher than DoubleA. However, when coding video image with the same frame number, XOR A can consume 0.1k bits less than DoubleA. That's to say, unequal error protection plan can get good balance between error protection and redundancy consumption.Because of the influence of geographical conditions, weather, the collision of multi-user and so on, radio channel tends to bring about biggish network fluctuation, which behaves the characteristics of non ergodic property and making an error easily. So it should take a measure of an effective error code protection. Aiming at these performance of the radio channel and the characteristic that the effects of the different parts'information in code streams produced by H.264 encoder are different, this article brings unequal error protection forward based on RCPC channel coding.Traditional channel coding methods implement error control channel coding with uniform rate to the code stream encoded by and outputted from source encoder. According to the fact that the different parts of code stream produced by H.264 source encoder have different impact on the reconstructed quality of images, this article adopts classified channel coding. For the bits of code stream produced by the encoder, which play an important part in rebuilding the video image, the channel coding with high level protection is adopted. It means that using more supervisory bits carries out the work of supervision and protection for the code stream produced by the encoder. While for the bits which is less important for the image reconstruction it uses the low-grade channel coding. The purpose is to implement unequal error protection mechanism for the code stream produced by the encoder and ensure the reliability of video transmission through the time-varying and error-prone radio channels in some degree.This article uses rate compatible punctured convolutional code (RCPC) channel coding as the channel coding in system design. Besides owning lots of advantages of convolutional code, RCPC can work according to the rate compatible punctured matrix. Choosing different rate compatible punctured matrix may produces the codes with different rates and changes the low rate codes into the high rate codes by rate compatible punctured operation. The approach is that deleting some coding bits selectively in the output terminal of convolutional encoder can produce high rate convolutional code by puncturing the convolutional code with 1/n rate and keeping the complexity of the low rate decoder with 1/n rate. It's very suitable for applying to the situation which needs variable rate channel coding. In the process of channel coding, it would be better to use the same RCPC with variable redundancy based on the convolutional code with 1/n rate and deleting different numbers'redundance to implement the protection mechanism with different grades, than to adopt the different channel coding for the bits with different importance level.The result is that, unequal error protection plan based on RCPC channel coding transmits video information through the time-varying, error-prone wireless channel. It's really an efficient error-resilience method. It has a better system performance than not using UEP scheme obviously. Under the condition of all kinds of channel SNR, PSNR Y in the decoding terminal can improve about 2db. It can ensure the reliable transmission of the video information and can also use the limited bandwidth resource of wireless channel efficiently.Two unequal error protection plans given in this paper both have their own advantages. Unequal error protection plan based on XOR A mainly restrains the decrease of the communication quality because of packet loss. While unequal error protection based on the RCPC channel coding is put forward aiming at the characteristic of time-variation, error-prone and limited bandwidth. These two methods both implement efficient error protection through the limited redundancy and get good balance between the error protection and redundancy consumption.
Keywords/Search Tags:H.264, Data Partition, Unequal error protection
PDF Full Text Request
Related items