Key Technology Study On Motion JPEG2000 Video Coding

The Joint Photo graphic Experts Group submitted a standard: Motion JPEG2000 formoving images in 2002. Motion JPG2000(MJ2) codes each frame in sequenceindependently using JPEG2000. The compressed bitstream is scalable. Compared toother video coding standard, Motion JPEG2000 has some advantages such as:robustness, easy editing and etc. It is widely used in digital cinema, digital camera,digital video, high-quality frame-based video recording and editing, remote surveillanceand etc. Seven Hollwoods giants selected Motion JPEG2000 as compression standard tobe used in digital cinema system in 2005. With support of entertainment and industrialcommunity, Motion JPEG2000 has promising future. The paper probes into MotionJPEG2000 key technologys in its application field.Firstly, the paper discusses variable bitrate control in Motion JPEG2000. Accordingto human visual system, sequence with consistent quality present higher performance.The paper proposes a rate control algorithm, based on rate-distortion function prediction,for non-realtime application. It also presents a variable rate control method for realtimeapplication. In the proposal methods for non-realtime application, only parts of framesin sequence are coded and their rate-distortion functions are recorded. For thoseuncoded frames, their rate-distortion characters are predicted according to knowninformation. Compared to previous algorithms, the proposed method achieve moreconsistent bitstream while acquiring same buffers for rate control. In the proposalmethod for realtime application, each buffer that stores a coded frame has acorresponding lookup table. Lookup table instead of discrete function is used to recordthe rate-distortion function. The lookup table is suitable for hardware implementation tofind the threshold. The method has constant delay and can be used for those consumerelectronics products which has constant resolution and compression ratio(such as digitalvideo and digital camera).Motion JPEG2000 code each frame using JPEG2000. The rate control algorithm inJPEG2000 coding is called PCRDopt(Post compression rate distortion optimization). Since the rate control is proceeded after the Tier1 coding,the encoder has to code allpasses and record all rate-distortion characters for each codeblock even though lots ofpasses will be discarded in the final codestream. It will causes great waste especially atlow bitrate coding. When the video has higher frame rate, too much wasting time inTier1 blocks realtime realiztion for Motion JPEG2000. The paper then focus on how todecide truncation threshold as soon as possible in Tier1 coding. A rate control method isproposed for Motion JPEG2000. It utilizes two features of MJ2: accurate rate controland rate-distortion correlation among frames of the same scene. It determines the initialcutoff threshold for the coding frame according to the previous frame's character andavoids encoding unnecessary coding units. We then add feedback scheme in theproposal method to further decrease the computation cost in Tier1. New cutoff thresholdis decided after each pass coding. With more codeblock coded, the cutoff thresholdcomes to the actual threshold gradually. The coding cost in Tier1 is decreased further.The proposed method is efficient at any compression ratio.Region of interest(ROI) is an important feature of JPEG2000/Motion JPEG2000.Dynamic ROI coding is often required in remote medical system and surveillance ones.In dynamic ROI coding, both the ROI region and its priority can be modified freely.Chapter 3 discuss the ROI coding in JPEG2000/Motion JPEG2000. Two dynamic ROIalgorithm, based on codeblock, DIRC and DWARC are proposed. The proposedstrategies produce a fully compliant JPEG2000 codestream. They allow transmission ofdifferent ROIs with different priorities and support dynamic delineation of them. Thedecoder the proposal algorithms required is compatible with JPEG2000 standard.TheDIRC rearranges truncation point for every codeblock in each layer. It assigns higherbitrate to ROI and lower bitrate to non-ROI and combines them to codestream. TheDWARC can achiev the accurate bitrate control in each layer. Firstly, the image isencoded as normal. Then it reallocates the bits allocated for ROI and background. Bythis way, ROI can achieve higher performance.Data may be damaged or missed while transmitted in internet or wireless channel. Chapter 4 is about the error concealment technique for Motion JPEG2000. Theinfluence of damaged subband depends on the subband frequency. Error in lowfrequency subband has the worst effect. For the worst situation: all date in LL subbandare damaged, we propose a reconstruction method. It choose a reference frame, which isreceived correctly and belongs to the same scene as the corrupted one. The Sobeloperator is applied to get edge image of reference frame and the corrupted one.According to motion estimation result for edge images, the corrupted frame iscategoried into frame with high motion, frame with part of motion and still frame. Thereplacement method is appied to reconstruct LL suband for still frame, while motioncompensation is applied to other frames. The paper also discusses the method'seffectiveness when other subband, as well as LL subband, occurs mistake. Theexperiments prove the method's validity.