视频中错误检测

3星(超过75%的资源)
所需积分/C币:10 2011-11-18 16:38:08 124KB PDF
8
收藏 收藏
举报

视频中的错误检测文档。主要针对H.264/AV压缩标准的视频流使用基于图像内容的方法进行检测。
which is consistent with the Cabac model. The For marker SE the new probabilit probabilities of Markers can vary in specific position to tM provide unequal crror detection. For example, svntax N(SE)+tM PSEn)+ elements in position A are more important, small probability P(,) markers are needed here to avoid misdetection. In less L+tM 1+ important positions. markers with relative high probability can be chosen, for sake of stream size saving Before adding markers, the compression ratio is given 2.4. Theoretical analysis by R slog, 1/H(p). I is the number of probability states of Since crror control schemes are gencrally achieved by syntax elements in the file, in CABAC, 1=63. After adding adding redundancy, the expansion of target stream size a marker. this becomes R'slog,l!H'(p). If the versus control efficiency needs to be estimated for measurement compression algorithm is effective enough, equalities can be As stated above, in the previous work of boyd et al. [1 used. Thus, the size of compression file after adding the an error detection scheme is proposed within arithmetic markers L+IM (L+(M)H'(p coding. The file redundancy and detective ability can be R og2 traded off by a reduction factor e. To add a %o redundant L The percentage of file expansion(Normalizing to -) hoose e 1+0.00693△ log,lL+tM I L -)/H(p) Differ from the general strategy on introducing a R reduction factor; our detection scheme by using markers can tM t give unequal error detection capability. Since precedence in olog(1+-)+p(senlog p(sen H(p) L CABAC Syntax elements vary, a worst-case design would lead to a prohibitive amount of redundancy tM tM (P(SE)+)logp(se)+.1 (4) In our scheme, if a precise and quicker detection is L needed, low frequency in Fig 3 can be chosen, thus means large probability state index. Video stream size will become Here, we can infer that stream size expansion can be considerable if rigor misdetection is needed. Any LPs controlled by marker numbers and marker probabilities as probability index can be chosen with regard to practical well. Trade off can be made based on real situation needs With this marker strategy, special markers are chosen to 3. EXPERIMENTAL RESULTS be embedded in CABAC, whose probability is pmar ker. If tror occurs in the encoded data. the probability of As stated above, smaller misdetection probability will misdetecting pms is the probability of still having the marker misdetection probability is shown in Table I. Our at the specific location, which is Parker. Clearly, the smaller experiment selects the LPS state index denoting by the pm ker, the lower the pm is. Since we use more than one pentagram in Fig. 2. Markers in this test are placed at marker in a single macroblock, pmi=p'muker, where t is the Position A and Position c with reference to Fig. 4 Forman"with QCiF resolution (176x 144)are encoded to number of times the markers encoded into cach macroblock produce the H.264/AVC test bitstream, JMIO I is used here However the size of the stream increases. Th as test source coder probability Pse of certain syntax element is N(SE)!L It is obvious that more markers placed behind different where N(SE )is the total number of times SE occurs in the syntax elements will generate more precise detection stream, and L is the size of the stream. By adding a marker However, stream size will become large. Thus Marker can after each block of size km, the new probability of be chosen according to the specific application as well as video stream characteristics. In test sequenceForeman occurrence of each syntax element in CABAC not used as a two markers are suggested to be implanted in position Aand marker is B respectively referring Fig. 4. In"Mobile"sequence, due PSE )=N(SE, P(SE to the high frequency information it possesses, one marker is L+tM tM inserted in position C as a complement. L Fig. 5 compares the bit-rate increasing on different where m is the number of blocks in the file. that is sequences. Markers are placed at position A, B and C(Fig 4) in all sequence. Sequences which possess more abundant K =L texture and complex motion are less vulnerable to bit-rate m=1 increasing, because marker information in them is miniature comparing to the original dense motion or texture Information 12.00 一 Foreman Table 1 Stream size versus misdetection probability 口 Mobile Marker Misdetection St Siz .e.Mather &.Daughter Probability state Probability expansion 8.00% 一米Bus Index 0.1205 0.0135 2.54% 600% 0.0031 4.03%0 7.29E04 5.39%0 3.52F-04 6.15 8.17E05 6.40% 63 2.54E05 10.32%0 To give a comprehensive demonstration of the detective 0.00 ability in this scheme, we adopt the multipath RayLeigh Fading Channel to simulate the real wireless transmission bability state index environment whose multipath profile is Brazil C with 6 Fig 5 Comparison on bit rate increasing ths Time selecti Doppler shift is 4 Hz. RS (204, 188) is utilized as the channel codes. Considering the Table 2 Marker Detection ability versus other scheme actual channel state. decoding ber is confined below 10 LPS Detected Error Detected Error In Table 2, we compare two detective methods under the Probability Decoding BER Location Offset Location Offset CABAC framework Scheme A in Table 2 is marker strategy, state index Scheme a Scheme B comparing to Scheme b('forbidden"symbol)[1]. Detected E-04 3 Byte 10 Byte error location offset is defined to evaluate detective abilit E-05 7 Byte 21 Byte The offset means the distance between the detected error 20 E-04 23 Byte 21 Byte location and the actual error location. Obviously, the smaller E05 5 Byte 8 Byte the offset is, the faster the detective speed it possesses 30 E-04 53 Byte 58 Byte Foreman'with QCIF resolution is used here as the test E-0 6 Byte stream. Results indicate that nearly under all LPS 40 E-04 4 Byte 6 Byte probability states and decoding BER simulated, the marker E-05 3 Byte 31 Byte strategy is superior to the reduction factor scheme in terms 45 E04 181 Byte 196 Byte of detective speed. Thats because Scheme B needs some E-05 2 Byte 12 Byte time to narrow down the range in order to discover the 55 E04 352 Byte 354 Byte former transmission errors F-05 I Byte 28 BVte 551 Byte 555 Byte 4. CONCLUSTION 4 Byte 172 Byte The proposed coding into CABAC uses context-based markers to determine errors. which demonstrates an efficient 6. REFERENCES performance in terms of flexibility and detection speed. This context-based error detection scheme can be used in 1 C. Boyd, J. Cleary, S.Irvine, I.Rinsma-Melchert, and cooperation with channel error control to guarantee I Witten, "Intergrating error detection into arithmetic transmission quality. Error tracking scheme can also utilize marker strategy for vidco transmission, which is combined coding, IEEE Trans. Commun., vo1.45, pp 1-3, Jan. 1997 [2 Jim Chou, Kannan Ramchandran, "Arithmetic coding with intra/inter-mode switching at source side Communications, vol 18, No6, pp 861-867, June. 200(epr based continuous error detection for efficient ARQ-based image transmission, IEEE Journal on Selected Areas in [3] George F.Elmasry, " Joint Lossless-Source and Channel Coding Using Automatic Repeat Request, IEEE Transactions on Communications, vol. 47, No. 7, pp953 955,July.1999 4 Detlev Marpe, Heiko Schwarz, Thomas Wiegand Context-Based Adaptive Binary Arithmetic Coding in the H.264/AVC Video Compression Standard, IEEE Transaction on Circuits and Systems for Video Technology 3,No.7,pp.620-636,July.20 692

...展开详情
试读 4P 视频中错误检测
立即下载
限时抽奖 低至0.43元/次
身份认证后 购VIP低至7折
一个资源只可评论一次,评论内容不能少于5个字
peggy_zhu 文档是英文的,很简短,不明白其中的细节。
2014-06-03
回复
u011188458 不是很好用
2013-12-09
回复
aqdjuly 基于内容的视频错误检测
2013-05-03
回复
您会向同学/朋友/同事推荐我们的CSDN下载吗?
谢谢参与!您的真实评价是我们改进的动力~
关注 私信
上传资源赚钱or赚积分
最新推荐
视频中错误检测 10积分/C币 立即下载
1/4
视频中错误检测第1页

试读结束, 可继续读1页

10积分/C币 立即下载