O-MOMS Supersampling routine

Supersampling function using Optimal Maximal-Order-Minimal-Support as kernel.

Description

MOMS (maximal-order-minimal-support) functions give the least number of supports for a given approximation order L. The stringent requirement of number of supports is critical to real-time signal processing, which is why sinc (the kernel that gives ideal reconstruction) is not used in practice. B-spline based interpolating kernels are usually used in spline interpolation. MOMS functions are constructed by b-spline functions. Here we provide an implementation of O-MOMS (optimal MOMS), which outperforms b-spline kernels of the same degree.

This implementation uses DTFT to compute the coefficients in the prefiltering step [Thevenaz 2000]. For boundary condition, we assume periodic, which is scheduled to be changed to mirroring in the next release.

Degrees of 0 through 5 are supported.

Related papers:

T. Blu et al.

Minimal Support Interpolators with Optimum Approximation Properties, ICIP 1998;
MOMS: Maximal-Order Interpolation of Minimal Support, IEEE Transactions on Image Processing Vol. 10, No.7, 2001;

Phillippe Thevenaz et al.
Interpolation Revisited, IEEE Transactions on Medical Imaging, Vol. 19, No. 7, 2000.

• MATLAB release: MATLAB 7.8 (R2009a)

[bmv]

The strange behaviour of the function.
If you make a resampling, and then make the binning and again resampling , there will be offset relative to the original series!
Example:
A - original
B-resempled and binned
max (A) index 70 106
max (B) index 69 105

At resempled images (128.128 -> 4096.4096)
max (A) index 2222 3361
max (B) index 2207 3345

[Meng Wang]

Hello,

Could you help me reproduce this weird thing? Thanks.

[bmv]

Ok. Simple example.
L=0:0.1:10;
S=sin(L);
Sr = momssupersample(S.', 4, 5);
Sb=imresize(Sr,0.25,'box');
plot(S-Sr)
plot(L,S,L,Sr)

zoom in

[Guillaume]

Hello,
He ask me the toolbox images, is it normal ?