Presentation is loading. Please wait.

Presentation is loading. Please wait.

PREDICTIVE SHAPE CODING USING GENERIC POLYGON APPROXIMATION Jong-il Kim and Brain L. Evans Circuits and Systems, 1998. ISCAS '98. Proceedings of the 1998.

Similar presentations


Presentation on theme: "PREDICTIVE SHAPE CODING USING GENERIC POLYGON APPROXIMATION Jong-il Kim and Brain L. Evans Circuits and Systems, 1998. ISCAS '98. Proceedings of the 1998."— Presentation transcript:

1 PREDICTIVE SHAPE CODING USING GENERIC POLYGON APPROXIMATION Jong-il Kim and Brain L. Evans Circuits and Systems, 1998. ISCAS '98. Proceedings of the 1998 IEEE International Symposium on, Volume: 5, 1998

2 Outline Lossy Shape Coding Using Generalized Chain Coding Generalized Predictive Shape Coding Simulation

3 Lossy Shape Coding Using Generalized Chain Coding

4 Lossy Shape Coding Using Generalized Chain Coding (con’t) Polygon approximation

5 Lossy Shape Coding Using Generalized Chain Coding (con’t) Vertex coding –Ri = Vi – Vi-1 –x_max_magnitude, y_max_magnitude –x_dynamic_range_indicator y_dynamic_range_indicator Example : V = (80,60),(105,66),(85,70),(70,63) R = (25,6),(-20,4),(-15,-7) x_max_magnitude = 25, x_dynamic_range_indicator = 4 y_max_magnitude = 7, y_dynamic_range_indicator = 2

6

7 Lossy Shape Coding Using Generalized Chain Coding (con’t) octant number containing the sign bit of x and y A flag indicating which component is the major component –Example : Ri(y) < Ri(x),Ri’s major component is the x-component

8 Lossy Shape Coding Using Generalized Chain Coding (con’t) If Ri(y) < Ri(x),Ri’s major component is the x-component –major_component needs (x_dynamic_range_indicator+1) bits –Minor_component needs min( y_dynamic_range_indicator+1,  lg|Ri(x)|+1  )

9 Lossy Shape Coding Using Generalized Chain Coding (con’t) Example : –Dynamic range –31~+31, Ri = (6,-3) –Header : x_dynamic_range_indicator = 4=>3 bits y_dynamic_range_indicator = 4=>3 bits –Ri in bitstream : Octant = 7=>3bits major_component (|x|>|y|,|x|=6) 4+1=>5 bits minor_component : min(4+1,  lg(6+1)  )=>3 bits

10 Generalized Predictive Shape Coding

11 Contour Motion Estimation Ic, Ir : current and reference contour set Nc, Nr : current and reference contour length A contour pel matching algorithm finds the offset vector (mv x, mv y ) to maximize the number of element in Ic ∩ Ir (mv x,mv y ) Comparison operation : 2*Nc*Nr times

12 Contour Motion Estimation Global motion estimation : all of the contour in a VOP Local motion estimation : single contour in a VOP Lossless or Lossy motion estimation : –A contour pel in current contour is within d max range from a reference contour pel

13 (a)(b) d max =1 (c)(d) d max =2

14 Coding of matched and mismatched contour position – Reference Index Based Coding Matched contour position –2D -> 1D –Reference contour length is Lr , just need  lg Lr  –Initial index is 1(the first pel during raster scan) , counter clockwise direction have increasing indices –For example (last page) Lr = 100 (V2,V3,V4,V1) = (30,38,45,58) Mismatched contour position –Generalized chain coding (octant,major,minor)

15 Simulation Ec : average bit for a contour pel

16 Simulation (con’t)

17

18 Conclusion Propose a new predictive binary shape coding method


Download ppt "PREDICTIVE SHAPE CODING USING GENERIC POLYGON APPROXIMATION Jong-il Kim and Brain L. Evans Circuits and Systems, 1998. ISCAS '98. Proceedings of the 1998."

Similar presentations


Ads by Google