The Study Of Video Object Adaptive Segmentation And Video Compression Coding

The vision is the most important way of obtaining external informationfor the mankind. It gives people the ocular and vivid image in visioninformation. The figures of the video information have a great deal ofadvantages that it can be recycled relaying, easy encrypting and have stronganti-interference ability etc. But it is very disadvantageous to store andtransmit a large amount of bits needed for the figure, it have alreadybecome one of the bottlenecks of hiding the mankind from obtaining andusing information effectively. A main purpose of the video code is base onthe premise that sure reconstructs quality to count and signify thepicture/video information for bits that tries my best to lack.The traditional compression code is set up on the basis of Shannon'sinformation theory, It is based on classical set theory to describes themessage source with the statistical probability model ,But it fails toconsider the subjective characteristic of the person who accepts information,Concrete meaning,important degree and consequence caused of theincident. The development course of the video code regards Shannon'sinformation theory as the starting point actually, and the course that isperfecting constantly.For a long time, on the basis of picture element method is being amajor method of the picture code all the time. It is setting out fromdispelling the relevantly and redundantly of picture date, the code entity ispicture element or block of picture element ,it take the display device as thelast link for the picture/video system, it has not considered the impact ofvision characteristic of human eye on coded picture.At the beginning of the eighties of the 20th century, people realizedsuch deficiency based on first generation of codes technology of the datastatistics, especially there is serious limitation in the video code of the lowspeed, so it is knew as the lower compressed coding method. But thecompression code international standard that adopting the technology of thefirst generation codes to be made has still obtained the universallyacknowledged immense success such as MPEG-1, MPEG-2 and JPEG, etc.The mid-1980s of the 20th century, relevant discipline rapiddevelopment and new developing discipline have injected new vigor intodevelopment of the video code constantly, at the same time the researchresults of the human vision physiology and psychological characteristicexpand people's vision. The entity in the picture code at this moment isno longer picture element or picture element block. But it is dividedaccording to its'content, the human eye is the final recipient of the pictureand video signal. The technology of the second generation code is notconfined to the frame of Shannon's information theory, but demand to fullyutilize physiology,psychological characteristic of the human vision andvarious kinds of the message source nature. The second generationtechnology is content-based, it remove the redundancy of the content frompicture and video signals, Among them, the method based on object iscalled the middle level compressed coding method, the method based onsemantics is called the high level compressed coding method. The secondgeneration technology based on content is the new generation video codingtechnology. This kind of method generally requires the pretreatment carrieson the picture, in order to cut the picture date apart according to the visionsensitivity, this is the most active field in the video coding at present. Mpeg₄ standard has used object concept during describinginformation for the first time. such as video object VO,audio object AO,these are a new leap. The so-called object can be visited (search for orbrowse through) and operated (clip and paste) in a scene, it can regard itslamination, form movement, etc as the basis while cutting object apart. Thisdescription method based on content is accord with people's psychologicalcharacteristic more than other, it can not only obtain super compressingperformance than existing standard based on picture element, but alsooffer various kinds of new function based on content interactive forapplication. Said from the code scheme, MPEG-4 is still a mix code that takes thesub-block as the foundation, this still has the very big disparity between itsoriginal intention and people's estimates. These indicate the research of thevideo code technology also has many work to need to do. The video object division and withdraws is the foundation for thecontent-based video application ,these applications include content-basedvideo retrieval, object-based video compression and edition, intelligentman-machine exchange etc, along with Mpeg-4 frame proposing, peoplestepped up to the video object division algorithm research. The goal of thevideo division is divide the object in the scenery from the same background.Each part of identical object has the consistent attribute. Generally speaking,only two aspects: spatial attribute and time attribute. These are the physicalbasis for all video division algorithm .The spatial attribute mainly is:Brightness, color, texture or other transformations statistics characteristics,for instance gradient picture, Para genesis matrix, histogram and so on.Generally has two kind of different considerations angles: Theregion(focuses to spatial attribute uniformity) and the edge (focuses tospatial attribute difference).Time attribute (movement attribute) mainlyperform for frame difference, light flow field or movement vector,according to the above may examine the frame changes (movement) theregion, as well as movement direction and size. The spatial division method may obtain the precise edge of the objectin the image. But the spatial division often has a bigger blindness, becausein the background static object is also divided, but finally also must bemerged together. In fact, the people often interest the movement objects,specially regarding the compression code, the object parameter ofmovement is extremely important. but we merely often can not obtain theobject's precise edge using the movement information, for instance usingframe misses, we may convenient, fast obtain the change part from twoframe, this difference template which obtains is not only incomplete butalso contains the background which revealed to be partial, therefore thedifference template can not direct be used as object template. We combine the inter-frame difference and image gradient. Anapproximate location for moving object can be obtained by using thedifference between two continuous frames, then the exact boundary isdefined according to the gradient of image, and finally an edge template isformed, from which regard minimum outer rectangle frame of object edgetemplate as the initial outline, then repeatedly carry on 'distanceadjustment',' outline low pass filter', 'move fixed point according to sendingline direction (stick decides image)','outline automatic fission' and so onstep, satisfy until restraining the condition. During the process ofdisappearing, it can divide and surround much object automatically,moreover because initial edge template is incomplete, must have a renewalprocess. The video object segment involves the video content analysis andunderstanding, these ties closely with disciplines such as artificialintelligence, image understanding, pattern-recognition and neural network,etc. At present the development of artificial intelligence is not perfectenough, the computers have not the ability of observe, dissert andunderstand image. Meanwhile the research of computer vision indicatesthat realize the image segment correctly must understand the video contentin a higher level. Therefore although the MPEG-4 frame already formulated,but until now there still have not the general effective method to solve thequestion of the video object segment, which is considered to be onechallenge difficult problem, content-based segment is more difficult. Video object plane (VOP) is the sample at a certain moment of thevideo object (VO),VOP is the key concept of MPEG-4 code.MPEG-4 adoptdifferent code tactics to different VO in the course of code: namely retainsthe detail and smooth to the compression code of prospect VO as far aspossible; adopt the high compressing rate code tactics to background VO,even refuse to transmit background VO and splicing it by otherbackgrounds at the end of decode. Such object-based video code has notonly overcome the block effect that the high compressing rate codeproduces in the first generation of videos code, but also it make user to bepossible with the scene interactive, thus it both improve the compressionratio, and realize content-based interactive ,these provide the broaddevelopment for video code. Along with the digital video code technology development, there aremore and more request and application of video transmission by IP networkwhich has the speed fluctuation and by heterogeneous network which hasdifferent transmission characteristic. Under this kind of background, theimportance of the scalable video code is outstanding day by day, itsapplication is extremely widespread and it have very high theoretical...