150 likes | 293 Views
Adaptive Context-Based Arithmetic Coding of Arbitrary Contour Maps. Armando J. Pinho IEEE SIGNAL PROCESSING LETTERS, VOL. 8, NO. 1,JANUARY 2001. Outline. Contour Map Differential Chain-Coding Transition Point Experimental Results Conclusion. Contour Map. DIFFERENTIAL CHAIN-CODING.
E N D
Adaptive Context-Based Arithmetic Coding of Arbitrary Contour Maps Armando J. Pinho IEEE SIGNAL PROCESSING LETTERS, VOL. 8, NO. 1,JANUARY 2001
Outline • Contour Map • Differential Chain-Coding • Transition Point • Experimental Results • Conclusion
DIFFERENTIAL CHAIN-CODING • r1 , c1 , s11 , s12,…, s1n1 , … , rk , ck , sk1 , sk2,…, sknk • (ri , ci) denotes a pair of image coordinates marking the start of a chain • sij represent the construction of the chains
DIFFERENTIAL CHAIN-CODING (con’t) • Two type of Contour map : Closed and Open • Closed Contours • There are four instructions of sij : • A (go-Ahead) • L (turn-left) • R (trun-right) • + (ramification)
DIFFERENTIAL CHAIN-CODING (con’t) • Example: • 0,1,L,R,+,A,A,R,+,A,L,L,+,L,A,A,L,R, 3,3,+,A,A,L,R,A
DIFFERENTIAL CHAIN-CODING (con’t) • Open Contours • An additional instruction is required indicating end-of-chain • A (go-Ahead) • L (turn-left) • R (trun-right) • + (ramification) • - (end of chain)
DIFFERENTIAL CHAIN-CODING (con’t) • Example: • 0,2,A,+,A,+,R,-,+,A,A,L,R,L,+,R,A,L,A,A,2,0,-
TRANSITION POINT • Four symbols {V,H,C,} =>
TRANSITION POINT(con’t) • At beginning of each row or column,the pen is up • While moving across a row,the pen changes state (up to down or down to up) if it encounters a corner or a horizontal transaction point (C or H) • While moving across a column,the pen changes state (up to down or down to up) if it encounters a corner or a horizontal vertical point (C or V)
TRANSITION POINT(con’t) • Adaptive CAE(context-based arithmetic encoder) compression methods : (a) Context are calculated in the domain of transition point (Ar-Tr),there are 46 possible contexts (b) Context are calculated in the domain of contour map (Ar-Ct), there are 28 possible contexts
EXPERIMENTAL RESULTS • “couple “ image into 9327 regions(55519 active contour elements)
EXPERIMENTAL RESULTS(con’t) 20 40 60 Number of active contour elements(x1000)
EXPERIMENTAL RESULTS(con’t) Number of active contour elements(x1000)
Conclusion • The proposed method behaves better than differential chain code or on the MPEG-4 shape coder. • Compression efficiency increases when the contour complexity increases.