色情報と知識処理に基づく車載カメラ映像のインデキシング

全文

(1)2006-AVM-54 2006/9/29. IPSJ SIG Technical Report. Indexing of On-Board Captured Images Based on Color Information and Knowledge Processing Mikio SASAKI Research Laboratories, DENSO CORPORATION 500-1 Minamiyama, Komenoki-cho, Nisshin-shi, Aichi, 470-0111 Japan Abstract. In this paper, we propose a method that generates image descriptions relatively easily by using color information and knowledge. processing. At first, for each block pixels of on-board captured images that have been stored in MPEG format, we have generated object. indices triggered by color information from decoded DCT coefficients. Secondly, by concurrently using three dimensional model based method and rule based reasoning, we have extracted three dimensional relative locations between major objects and scene features. Those results have been used to generate final image descriptions in XML format automatically. Moreover, we have confirmed through the experiments that it is possible to conduct image retrieval by using natural language like queries, on the basis of matching operations between the image descriptions above and query profiles.. Keyword. Indexing, DCT, model based, on-board captured images, image retrieval. Ell Fig.l System architecture.. - 45 -. (9).

(2) (a) Macroblock structure. (4) Y compoaeot values. (c) Color indexed image. (b) Decoded image in MPEG-1 format. (e) T? compoceot values. (?) V component values. Fig. 2 Analysis of YUV component. 2.2. -f. 2.3.. CttTttftfcfcft. Evening (rain). 03. Nighttime (rain). ^^y-^V^TM^SkO^]. Hg.3 Examples of color indexed images. -46-.

(3) z'*- n—v -f. XML. .. K: YUV DCT. L * V^. o T MPEG-1 x-. YUV. 3.2.. 2). 3) .. MPEG-1. (16X16. 3.3.. (in, «), .3. IX--8mll. ^K. I Y°'I.2mijX«2wi. I X-!2. (X.Y.Z) (Oni. Om,- IOrn,J. Fig.4 Solid structure model of road scene.. -47 -. fc.

(4) &• 8'* 4.2.. ,. m 8(a). ntafitt ^JWBAT.meon far. iwj £= Yth. nrfr. (1). w* fa # MBK(mr,mc)<E>. ,. MPEG-1. (16X. ♦4. (8X »%&#*?!. Jmr,mc)=(l/4)ZYBLKjnean(mr,mc,i). (2). Fig.5 Distance information estimated from road structure model.. 06 Fig.6 Color indexed image.. C =. cfl. +. I c. (4). Fig.7 Three dimensional display of depth.. -48-. -1. <.

(5) (a) Y component. (b) U component. (c) V component. (e) Three dimensional display of c. Fig.8 Correction of depth on the basis of sky information. 5.2.3.. tt ". ,. 5.2. 5.2.1.. z, £. MMT. IE@T, 9).. s.2.2.. Fig.9 Area definition on image plane.. -49-. DC.

(6) AC (7). GAC =. i: GyflJ/GH[lJ. 7. Fig. 10 Time series histogram regarding red regions (nighttime). (9) h°OvO. 7. Gy[lJ=. AC. GAC £ 7.0. 5 \tM\Z Y0,Y,.Y2,Y3,U,V7 U. 7. 7. Z. X. h"0. v°0. (1/64). 7. (10). (5). Qac = Z SAC[l]. SAC[l]~. 7. 0/64). &ac'.. 1=0,1. (8). lB,(h,v)/ - /B,(0,0)/. (11). (6). io. f« RN2). RN3). L5. RED. Fig.ll TTifee dimensional expression of estimated results.. RE2). -50-.

(7) 7.. (b) Qutiy <tcsctiptk)fi. XML. fc,. (o) Scene description. (c) Rdrtevtd pictare. (DV Fig. 12 Description and retrieval of scene. PC±TMPEG-1. RE3). 7.2. 3. 6.2.. 2{Wk2lZ0(i,ND(k),n)-Z0(i,ND(k),n-l)l} ik=0. (12). i-0. 7.3. ^T-. 12(a)K:*f.. Tablet Experimental results of image retrieval.. -51 -. d tl.

(8) [1] ISO/IEC 15938-3 Information technology - Multi. media Content Description Interface Part 3: Visual. [2] ISO/IEC 15938-5 Information technology Multimedia Content Description Interface Part 5: Multimedia Description Schemes. [3] B.S.Manjunath et al., "Introduction to MPEG-7", WILEY, 2002.. (13). sin.. r-sKi—a,. vkmi~\i},. Bixi/. U=.!7Z, 2004^6^215^, pp.105-129. (14). ar=NP/(NP+NQ). (15). fl 24 0-^,. un^X, 2005^ 10. [6] "Information technology - Coding of moving pictures and associated audio for digital storage media at up to about 1,5 Mbit/s Part 1: Systems", ISO/IEC11172-1 First edition 1993-08-01.. (16). = 2/(l/ar+ 1/atp). pp.101-119.. [7] "Information technology - Coding of moving pictures and associated audio for digital storage media at up to about 1,5 Mbit/s Part 2: Video", ISO/IEC 11172-2 First edition.. 2 ?]• — &/®<D MPEG-1 ^>h d-h, J74-D-11-8, pp. 1497-1505, Aug.1993.. \Z7Tsir.. ." ,1997 *m (ITE'97),p.381-382,1997^7£.. [10] M.Sasaki, A.Toyoda, "3D Motion Estimation of a Moving Object using Model-Based Techniques", ITS95 pp.1075-1081, November 1995.. i^T 1/10 ^. ,. IiT« 1/100. £tz, ^^. Ltc.. (Dec.1990) .. -*T,. ,. Vol.31, No.12, pp.1743-1754. [12]Youiti Horry, Ken-ich Anjyo and Kiyoshi Arai, "Tour into the picture", Proc. SIGGRAPH'97, Los Angels, California, pp.225-232, August 3-8, 1997.. ^m,. ii,. &=?, u>®m®®&. M", ft^ft* MVE 2001-148 (2002-3). ]. tb%. -. : IF-THEN ;i/-. -6 3. A (PCSJ • IMPS98). 1-1.4, pp.9-10, Oct.26-28, 1998.. ITS2004-22, Vol.104 No.325, pp.35-42, 2004. [16]*. &ffl, $|^ffi3, "i««l*^JJX WJS, 2002. , "I T 9. MPEG-1. ", 2006 (JSAI 2006) 1E2-2, 2006 ¥ 6 fl .. 2 ^7C DCT. -52-.

(9)