The Improvement Of Motion Estimation Algorithm In H.264 Video Coding Technique

Video compression coding technology is developing together with internet andmultimedia service; it plays an important role in the video communicationservice such as DTV, HDTV, visual telephone, conference telephone andmultimedia, etc. Data compression is based on the principle that there are variousredundancies in the original data. By eliminate intra-frame space redundancy andinter-frame time redundancy, the video coding technology can realize the datacompression.As the most promising standard of dynamic video compression, with it'sexcellent compression performance and friendly network affinity, H.264 servesa wide range of applications, such as, digital television broadcasting, real-timevideo communication, multimedia communication and delivery via network.Based on H.263, H.264 is the latest low-bit rate digital video compressionstandard constitutedBy the ITU-T ( International Telecommunications Union).Compared with itsprior opponents, H.264 has much better quality owing to the adoption ofadvanced coding techniques such as integral DCT, context-based adaptivearithmetic coding (CABAC), intra prediction, 1/4 pixel accuracy motionestimation and seven kinds of matching blocks. So H.264 has a much broaderforeground in applications such as wireless communication and videotransmission over the networks. Therefore,how to reduce its computationalcomplexity and acce1erate the execution speed while maintaining theoutstanding compression performance is a crucial problem in the H.264applications.In the video coding and processing system, motion estimation (ME) andmotion compensation (MC) play a very important role. They can eliminate the inter-frame redundancy and improve the performance of the video coder. On onehand, the accuracy of motion estimation directly affects the efficiency of thevideo coder. It eliminates the inter-frame correlation and reduces the demand ofthe store memory and the bandwidth for transmitting greatly. On the other hand,the motion estimation costs most of the coding time of the encoder, therefore thecomplexity of the encoder lies on the complexity of the motion estimationalgorithm.Through studying each part arithmetic of H.264 standard and analyzingH.264 encoder,we can conclude that motion estimation is still one of the coretechniques of video coding of H.264, and in whole process of video coding,motion estimation is one of the most time-consuming parts,so how to increasethe efficiency of motion estimation and look for a faster and effective motionestimation algorithm has become a hot research hotspot at present. This paper isdoing research on the improvement of the motion estimation algorithm in H.264.This paper mainly studies the critical technique of H.264 based on therealization of H.264 coder and emphasizes the algorithm of motion estimation.The paper points out two kinds of motion estimation methods that are fit forH.264. Detailedly, the work includes:This paper first introduces different kinds of video coding standards, thebackground of H.264 and the features of H.264.Then the basic knowledge of video coding, the realization and the criticaltechnique of H.264 coder are discussed. Similar to the early standard, H.264 isalso based on the block and its main function modules including prediction,transform, quantization, entropy coding have not essential change. The changelies in the details of every modules and the change of these details is also thecritical technique of H.264.For the motion estimation consumes 60 per-80 per of the total amount ofthe coder, so the third part studies the block motion estimation. According to the previous motion estimation algorithm, the paper presents the motion estimationscheme of H.264 aiming at the high accuracy and small calculating.At last, through the research on large quantities of information and theoreticsand related testing, in accordance with the disciplinarian of spatial distribution ofmotion vector, this paper points out two kinds of fast motion estimationalgorithm based on template: the first one is triangle-small cross algorithm basedon predicting (TSCS), it introduce two kinds of new template and make themexerted in a parallel way, and has predicted the initialization; the second isparallel cross algorithm based on direction (OPCS), it fits the cross templatestogether and breaks them up by making use of the character of direction of thebest point's distribution, and takes the advantage of parallel search.Realize each algorithm by programming based on its basic idea. Use each oneon a series of testing sequences on JM8.6 which is a reference procedure modelof H.264 standard, comparing with some classical algorithms similar to it withthe same parameters. The measured data shows that: TSCS improve its searchingspeed observably, and maintain good image quality; For OPCS, there are certainadvantages in speed and the image quality dose not decline. Its integratedperformance is excellent and stable. In the end, compress a sequence with thetwo algorithms separately and compare the images with the original one. We cansee that the effect is petty good.The algorithms proposed in this paper both prove searching speed greatlyunder the premise of guaranteeing image quality. Under the trend of theapplications of H.264 are more and more extensive, excellent fast algorithm hasinestimable value and future prospects. Of course, these algorithms are onlybased on template. If we can combine them with other methods increasingsearching speed, such as halfway-stop technique and search window updateddynamically, and so on, it will be able to further improve the speed on the basisof them; at the same time as the part of motion estimation in H.264 adds to multi reference frame and multi-modal selection techniques, there is a great part of theoptimization of space. These are the directions to further reducing motionestimation computing and improving the speed of H.264 coding.