tdinterp.c File Reference

Roussos-Maragos interpolation by tensor-driven diffusion. More...

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <fftw3.h>
#include "basic.h"
#include "finterp.h"
#include "tdinterp.h"
#include "conv.h"

Include dependency graph for tdinterp.c:

Go to the source code of this file.

Functions
int	RoussosInterp (float *u, int OutputWidth, int OutputHeight, const float *Input, int InputWidth, int InputHeight, double PsfSigma, double K, float Tol, int MaxMethodIter, int DiffIter)
	Roussos-Maragos interpolation by tensor-driven diffusion. More...

Detailed Description

Roussos-Maragos interpolation by tensor-driven diffusion.

Author

Pascal Getreuer getre.nosp@m.uer@.nosp@m.gmail.nosp@m..com

This program is free software: you can use, modify and/or redistribute it under the terms of the simplified BSD License. You should have received a copy of this license along this program. If not, see http://www.opensource.org/licenses/bsd-license.html.

Definition in file tdinterp.c.

Function Documentation

int RoussosInterp	(	float *	u,
		int	OutputWidth,
		int	OutputHeight,
		const float *	Input,
		int	InputWidth,
		int	InputHeight,
		double	PsfSigma,
		double	K,
		float	Tol,
		int	MaxMethodIter,
		int	DiffIter
	)

Roussos-Maragos interpolation by tensor-driven diffusion.

Parameters

u	pointer to memory for holding the interpolated image
OutputWidth,OutputHeight	output image dimensions
Input	the input image
InputWidth,InputHeight	input image dimensions
PsfSigma	Gaussian PSF standard deviation
K	parameter for constructing the tensor
Tol	convergence tolerance
MaxMethodIter	maximum number of iterations
DiffIter	number of diffusion iterations per method iteration

Returns

1 on success, 0 on failure

This routine implements the projected tensor-driven diffusion method of Roussos and Maragos. Following Tschumperle, a tensor T is formed from the eigenspectra of the image structure tensor, and the image is diffused as

In this implementation, the fast 5x5 explicit filtering scheme proposed by Weickert and Scharr is used to evolve the diffusion. The diffusion is orthogonally projected onto the feasible set (the set of images for which convolution with the PSF followed by downsampling recovers the input image). Projection is done in the Fourier domain, this is the main computational bottleneck of the method (approximately 60% of the run time is spent in DFT transforms).

Beware that this routine is relatively computationally intense, requiring around 2 to 20 seconds for outputs of typical sizes. Multithreading is applied in some computations if compiling with OpenMP. Multithreading appears to have little effect for smaller images but reduces run time by about 25% for larger images.

Definition at line 694 of file tdinterp.c.

Here is the call graph for this function:

Here is the caller graph for this function: