Sunteți pe pagina 1din 17

Comparing 3-D Interpolation Methods

Binu Enchakalody January, 12 2010

Interpolation Methods
Nearest Neighbor Tri-Linear Cubic-Keys Clamped Cubic Spline Catmull-Rom Spline Cubic B-Spline

Methods
Nearest Neighbor
Picks the intensity of the nearest x,y and z pixel within a 2 x 2 x 2 neighborhood

Tri-Linear
Weights an interpolated intensity value, based on the distance from the nearest x,y and z pixels within a 2 x 2 x 2 neighborhood Vxyz = V000 (1 - x) (1 - y) (1 - z) + V100 x (1 - y) (1 - z) + V010 (1 - x) y (1 - z) + V001 (1 - x) (1 - y) z + V101 x (1 - y) z + V011 (1 - x) y z + V110 x y (1 - z) + V111 x y z

Piecewise-Cubic
Basic Algorithm

[ t0 t1 t2 t3] x Basis Matrix x [a-1 a0 a1 a2] , where 0 < t < 1

Cubic-Keys

*Cubic Convolution Interpolation for Digital Image Processing - R.G. Keys C2 Continuity (curvature)

4 Cubic Kernel Dimension Separable

Catmull-Rom Spline

Different Basis Matrix

Piecewise Cubic-Spline
Clamped Cubic-Spline

C2 Continuity (curvature) 4 Cubic Kernel Dimension Separable

General Equation

Piecewise Cubic-Spline
B-Spline Interpolation
*A parallel B-Spline fitting algorithm F.Cheng, A. Gosthashby

Uses Control Points defined by the neighboring intensity pixels

Calculating Control Points, C


Parallel Algorithm
A parallel B-Spline fitting algorithm F.Cheng, A. Gosthashby

Original

Volume Center slice

Nearest Neighbour

Tri-Linear

Cubic-Keys

Catmull-Rom Spline

Clamped-Cubic Spline

B-spline

Analysis
Rotated a 151 x 221 x 50 Test Matrix, 6 times in increments of 60 degrees Along the Z-axis Along the X-axis (permuted) Recorded computation time for interpolating the data-set Error Analysis on a 23 x 29 x 11 sub-matrix Calculated MSE, Mean, Standard Deviation

Error analysis on test matrix, after 6 rotations of 60


- along Z-axis (top), along X-axis (bottom)
400 350 300 250 200 150 100 50 0 MSE 19.63 362.86 Nearest Tri-Linear Cubic_keys Catmull-Rom Clamped Spline B-Spline

18 16 14 12 10 8 6 4 2 0 Mean Std. Dev

Nearest Tri-Linear Cubic_keys Catmull-Rom Clamped Spline B-Spline

400 350 300 250 200 150 100 50 0 MSE 18.52 Nearest Tri-Linear Cubic_keys Catmull-Rom Clamped Spline B-Spline

18 16 14 12 10 8 6 4 2 0 Mean Std. Dev

Nearest Tri-Linear Cubic_keys Catmull-Rom Clamped Spline B-Spline

Approx. computation time@ rotation(MATLAB profiler)


700 600 500 400 300 200 100 Nearest Tri-Linear Cubic_keys Catmull-Rom Clamped Spline B-Spline

650

283 140 140 3 6 Time (s)

S-ar putea să vă placă și