Wang Junsheng
1 Introduction
The H.264 standard is a new generation of low-bit rate video compression coding standards for the video coding expert group of MPEG and ITUs. Compared with past standards, it adopts more advanced technologies such that the H.264 standard coding can be used in the same rate of code. At the same time, due to the high efficiency compression coding of H.264 new standards, the compressed video stream is more sensitive to channel errors during transmission, even if a single burst error, it is also possible to seriously interfere with normal decoding of the receiving end, resulting in recovery video quality A sharp decline. To this end, the new standard of H.264 uses a variety of encoding techniques for enhancing the compressed video flow anti-error error to ensure the quality of the recovery of the video stream, this article summarizes this.
2H.264 Video Codec Standard Baby Amnotation Technique is the same as other video codec standard H.263, MPEG-2 and MPEG-4, H.264 standard is also enhanced by combined applications of multiple anti-anemosis technology. The video stream is robust to the error in the error, the loss of multiple environments, such as wireless and IP channels. Compared with H.263 standard, the anti-mistake technology used in the new standard of H.264 can be divided into three categories, and one is an anti-obnegant technology that is highly efficient and technically mature in the old standards directly adopted by H.264. Such as image segmentation, reference image selection and other technologies; the second class is an anti-anegic technology that improves better application in H.264 standards, such as intra coding, data segmentation; third category is based on H.264 standard New video compression anti-error technology. Three new anti-compaute techniques are mainly used in H.264: parameter set, flexible macroblock arrangement (FMO) and redundant film technology.
The first type of anti-Comparative technology has a more detailed description in many existing literatures, which is not described here. This article is only discussed in the last two types of anti-challenge technologies mentioned above.
2.1 intra coding
Based on macroblock, Slice or image is used to overcome the influence of the reference image drift due to the error due to the error, this anti-anemosis technology is very good in the standard H.263. The application, but in H.264, intra-code technology is also given two new features.
(1) intra prediction
H.264 Allows the intra macroblock prediction, and can even use the constrained intra PREDICTION flag to identify whether the intra macroblock prediction is used to identify the intra-intra-intra PREDICTION FLAG. When the limit frame predictive encoding flag is set, it is indicated that the prediction of this method is not used, and the synchronization characteristics of the intra information is restored. In a wireless environment in a wireless environment discussed in detail in Reference [3], it has achieved good results by setting a restricted intra prediction coding mark.
(2) intra enchannels and IDR films
Two types of intra code, the new standard H.264, including only an intra-frame encoding macroblock: intra enchannels, and IDR (Instantaneous Decoder Refresh). Wherein, the IDR piece is only used to form a complete IDR image, that is, all the pieces in the IDR image must be an IDR slice, and one IDR piece can only be part of the IDR image. At the decoder end, after the decoded IDR image is decoded, the decoder immediately identifies all reference images as "not used as a reference". Thus, when subsequent images are decoded, the limbs do not refer to any image in front of the IDR image. The first image of each video sequence must be an IDR image. The IDR image has stronger emphasis characteristics compared to images that contain only intra-encoders. It should be noted that since the image referred to by H.264 is used in this intra-reference image, even if the encoded image and all subsequent images are transmitted even if this intra encoded image is transmitted, it cannot be eliminated. Image drift caused by the error diffusion.
2.2 Data segmentation
Data segmentation is an efficient anti-mistake technology. Typically, all symbols in one macroblock are encoded in a single bit string for constituting the sheet. The data segmentation generates multiple bit strings for each sheet, ie multiple data blocks, and symbolize symbols that are semantically relatively related to each other and closely linked in a separate data block. In view of the importance of information, the quality of the different data block is used to ensure the quality of the recovery video. In H.264, three different types of data block: header information, intra data blocking and?
(1) Head information
The header information contains the type, quantization parameter, and motion vector of this macroblock. This data block is the most important if it is lost, even if other data block is properly received, it will not be available, so the header information is given to the maximum protection after the resolution. This data block type is called a class block in H.264.
(2) intra data block
The intra data block is also referred to as a B-class block, which carries the intra CBPS and intra-inflated coefficients. Class B block requires the Class A block of the corresponding piece to be available. Contains synchronization between the elbow device, correspondingly, mistant; χ χ 钋 钋 A 硗猓 盇 盇 分 分 分 分 分 类 类 分 分 分 分.
When using data segmentation, the source encoder places different types of symbols into three different bit buffers to generate three types of data blocks. At the same time, the size of the film is adjusted to ensure that the package smaller than the MAXIMUM TRANSFER Unit when the maximum data block is packaged. It is also for this reason, in H.264, is executed by the source encoder to perform data segmentation rather than NAL.
In the decoder, all data blocks should be effectively used for image reconstruction. In particular, if only the intra or is only lost content information, it can be masked well with previous decoding frames.
2.3 parameter set
The parameter set is usually applied in all H.264 bitstreams, which is extremely important, and it will affect a large number of VCL and NAL units, which cannot be correctly decoded even if it can receive correctly. Two types of parameter sets were used in the H.264 new standard.
(1) Sequence parameter set, including all information related to the image sequence (defined as all images between two IDR images), applied to the encoded video sequence.
(2) The image parameter set contains the related information of all the sheets belonging to the image, which is used to decode one or more independent images in the encoded video sequence.
After the parameter set of multiple different sequences and images is stored correctly after the decoder is properly received, the encoder selects the appropriate image parameter set by reference to the storage location of each encoded sheet head, the encoder selects the appropriate image parameter set, and the image parameter set is included. 1 sequence parameter set to use and refer to.
Flexible use of parameter set greatly enhances the anti-comment capability of codec. In the wrong tendency environment, the key to using the parameter set is that when the corresponding VCL and the NAL unit reaches the decoder, ensure that the parameter set has reached the decoder in time. The most commonly used means are repeatedly sent to increase the probability of reliable data. In this case, the typical application is the transfer of the parameter set and a VCL NAL share 1 channel. In addition, the parameter set can also be transmitted separately in the band.
Due to the use of a reliable transmission mechanism and better performance, the parameter set is timely and reliably served the decoder end to ensure correct decoding of the corresponding VCL and NAL units. However, this approach requires additional channels, and a reliable transmission mechanism. When conditions permit, apply this transmission mode to enhance the anti-commencement capability of the codec, but is limited to the status quo of network resources, and more Many is to transmit parameter sets in the first way.
2.4 Flexible macroblock arrangement
The flexible macroblock arrangement (FMO) allows macroblocks to each piece in an organization that is different from image scanning sequence, in which case each macroblock is fixedly assigned to one piece according to the macroblock configuration diagram, The macroblock is encoded in the scan order, and each sheet is transmitted separately. If a piece is lost during the transmission, you can use other correct receptions, including the film that is adjacent to the macroblock in the lost sheet to make a valid error mask. Since each piece is independently decoded in the decoder, the spread of errors is effectively suppressed, and the ability to disable the decoding is improved. It is assumed that the image is sufficiently divided into two sheets, and all macroblocks in this image are configured to sheet 0 or sheet 1, where the white macroblock belongs to the sheet 0, while the gray macroblock belongs to the sheet 1. If in the transmission, the package containing the information of the sheet 1 is lost, since each macroblock in the lost sheet 1 is dispersed adjacent to the macroblock in the other piece 0, and the sheet 0 contains a large number of sheets 1 macroblocks. Related information, using it, you can make a valid error mask for lost pieces. Tests show that in video conferencing applications, CIF-sized images, even if the loss rate is 10%, the visual effect of reconstructing video after using FMO can only confirm that it is a refactoring video after loss. The cost of using FMO is its lower coding efficiency (because it destroys the prediction mechanism between adjacent macro blocks within the image), and in high optimization environments, due to large predictions, large Delay. FMO has a variety of modes, from rectangular patterns to rules, distributed patterns or completely random distribution patterns. .5 redundancy
Redundant sheet technology, simply, in the same bitstream, the encoder is placed in addition to the macroblocks that have encoded in the sheet, and the same macroblock is placed simultaneously. Among them, the influence of the unsurable sheet caused by the error is overcome by using one or more redundant representation of the macroblock. This technique is different from the redundant transmission, such as a package replication. In the package copy redundant transmission, the copied package and the copy package are exactly the same, and in the use of redundant pieces, the redundant sheets use different coding parameters to encode, thereby forming different representations to the same macroblock. For example, the main miniature sheet can be quantified using a smaller QP to quantify, so there is a good reconstructed image quality. Secondary redundant films can be quantified using a larger QP to quote, thus using fewer bits, and the reconstruction quality is rough. When the image is reconstructed in the decoder, if the main redundant piece is available, only the main redundant sheet is discarded using only the main redundancy. The remaining redundant pieces are used to reconstruct the image when the main redundancy is not available during the transmission process. This use of different coding parameters of the primary redundancy block makes the redundancy technique to maximize the quality of the reconstructed image, especially in the mobile channel with a misdiscipient Or in the IP channel environment, redundant block technology can significantly improve the maintenance quality of the reconstructed image.
3 Conclusion
H.264 is a new generation of low-bit rate video compression coding standards established by ITU-T and MPEG, which contains rich anti-ambigum technology, providing real-time, efficient transmission for compressed video streams. Guarantee. Analysis and research on new standard anti-mistakes can better utilize these technologies to improve the quality of end-to-end communications, and have extremely important roles and significance for the development and development of wireless and IP video communication products.
