| In Recent years, the development of multimedia leads to the rapidly increasement of multimedia data. Image data is the most enormous data among all types of data. Because of the image data is very huge, we can not solve the problem of storage and network transfers depend on increase disk space or enlarge network bandwidth simply. So, we have to compress image before it been transfered. Furthermore, we need real-time compress image data in some special occasions, such as net meeting, real-time monitor. In this article, we studied a method of how to realize the MPEG-4 on BF533 to realize the real-time compress image data based on the characters of ADSP-BF533. MPEG-4 is an international standard that was established by Moving Pictures Experts Group, one of ISO groups, in 1999. It is a stand that fits low rate transfers. MPEG-4 has many advantages in image compress aspect. First, it is based on content and has strong random storage ability. Second, it can offer high compression ratio. Third, it has diversiform storage and access methods. Furthermore, MPEG-4 has a little effect on the quality of pictures while it offers high compression ratio, so it is the reason that MPEG-4 has been used widely. Because of the technology of image segmentation is not mature enough that it is impossible to realize the alternation based on content. In fact, we use the image segmentation technology based on rectangle in this article. There exit strong time relativity and space relativity among image sequences, so we can use some arithmetics reasonably to reduce time redundancy and space redundancy effectively. These arithmetics include move estimation, move compensation, direct current and alternating current prediction, DCT transform, quantization and Variable Length Coding, etc.The move estimation and the move compensation are important parts in reduce picture's time redundancy. They use the character of time relativity of image data. They use reference frame to predict current frame, and they only encoding the difference of the two frames. So it reduced the encoding data efficiently and it plays an important part in reduce time redundancy. On the other hand, the arithmetics of direct current, alternating current prediction, DCT transformation, quantization and Variable Length Coding use the character of space relativity to deal with data and reduce space redundancy. Digital signal processing (DSP) is one of the most powerful technologies in today's world. It relates to many subjects, and it has been used in many scientific and project fields, such as communication, medicine imaging, radar, sonar, high fidelity music rebuild and petroleum prove, etc. And it also has been used in image process. We may have special needs in many situations, such as video monitor and net meeting. In these situations, we need real-time handle image data. Because of DSP has high frequency and low price, it has been used widely. It especially fits for low power, little size and fast network video equipment and wireless handset equipment. The excellent performances and low price of Blackfin series DSP of ADI Corporation makes it focused more and more attention. The series DSP of Blackfin adopt the micro signal architecture (MSA) exploited by ADI corporation and Intel corporation. It is a combination unit with the advantages of double MAC engines, concision correctitude dictate collection that like RSIC microprocessor and single instruction multi data (SIMD) multimedia. It is a sort of single instruction collectivity system structure. The MAC of Blackfin series DSP can accomplish an multiplication operation of 16 bits multiply 16 bits in one clock cycle and adds the result to forty-bit accumulator. The chip was made according the integrate circuit techniques of 0.18/0.13um, its tiptop work frequency can reach to 600MHZ and its peak operation ability can reach to 1200MMAC. So it can satisfy the real-time need of multimedia application for MPEG-2 and MPEG-4 absolutely. In this article, we realize the image data real-time compression mainly by the followed several steps. First, we need debug c program on PC successfully. For predigest, we only set some basic parameters when setting parameters, such as input file name, output file name and picture size. So we can find the core parts of this c program in order to pick them up and used in DSP. And we can do some optimizations such as turn some constants that often been used into global constants. Second, we should debug the program in DSP after it has been debugged in PC successfully. Therefore, we should build a "*.ldf "file to distribute the storagespace of DSP in order to avoid address overflow when the program is been debugged. Because of the storage space of DSP is limited, so we can not storage all data in internal memory, so we have to leave some tables in SDRAM. Third, do some optimize based on ADSP-BF533. In this step, we mainly do some optimize on some functions been used frequently. We can use assemble language instead of c language and also can we use some special multimedia statements. We should use parallel clauses as possible as we can. Generally speaking, one sentence is formed by one operation clause and two storage clauses and they are parallel operated. By doing so, the need of pipeline can be maximum satisfied and it can enhance encoding speed. Furthermore, we must set hardware reasonably. In this article, we use PLL set system clock and core clock in order to enhance system capability as possible as we can while the need of hardware been satisfied. After those three steps, we can basically realize the image data real-time compress based on ADSP-BF533. At the end of this article, we use some standard test sequences to do some tests and through the results of these tests we believe that it is possible to realize image data real-time compression based on ADSP-BF533. In this article, we compress image sequence about fifteen frames per second, although it does not meet the need of twenty-five frames per second, but it is relative reached that target, I believe we can reach that target finally via optimize the program further and make further use of the advantages of hardware.
|
PDF Full Text Request