pbellrescale

Performs a rescaling of image using the Bell algorithm.

Synopsis

pbellrescale zoomx zoomy zoomyz [im_in|-] [im_out|-]

Description

pbellrescale uses a convolution kernel to interpolate the pixel of the input image im_in in order to calculate the pixel value of the output image im_out. The interpolation consists in weigthing the input pixels influence on the output pixels. The weights are relative to the position of the output pixels and are given by the Bell algorithm:

       | 0.75-sqr(x) if -0.5<x<0.5
B(x) = | 0.5 * sqr(|x|-1.5) if -1.5 < x <1.5
       | 0 otherwise

For example, if the image is scaled by 3, then each output pixel is:

for i in [-2, 2]
  for j in [-2, 2]
      im_out[p.y][p.x]] += B(i*scalex)*B(j*scaley)*im_in[p.y*scaley+j][p.x*scalex+i]

To rescale region map or graph, use the operator prescale.

Parameters

zoomx, zoomy, zoomz are positive real values.
- if a zoom factor is > 1 then the image is enlarged along the related axis.
- if a zoom factor is < 1 then the image is shrunk along the related axis.
(zoomz is ignored for 2D images but must be given).

Inputs

im_in: an image.

Outputs

im_out: an image of the same type as the input image.

Result

Returns SUCCESS or FAILURE.

Examples

Enlarges the tangram.pan 2D image by a factor 2:
```
   pbellrescale 2 2 0 tangram.pan a.pan
```

Shrinks the tangram.pan 2D image by a factor 2:

   pbellrescale 0.5 0.5 0 tangram.pan a.pan

C++ prototype

Errc PBellRescale( const Img2duc &im_in, Img2duc &im_out, const float zoomy, const float zoomx );

Version française

Retaille d'une image par l'algorithme de Bell.

Author: Régis Clouard