SoRotateProcessing2d Class Reference
[GeometricTransforms]

SoRotateProcessing2d image filter More...

#include <ImageViz/Engines/GeometryAndMatching/GeometricTransforms/SoRotateProcessing2d.h>

Inheritance diagram for SoRotateProcessing2d:

Public Types
enum	PrecisionMode { NEARBY_PIXEL = 0, INTERPOLATION = 1 }
Public Member Functions
	SoRotateProcessing2d ()
Public Attributes
SoSFImageDataAdapter	inImage
SoSFVec2i32	rotationCenter
SoSFFloat	rotationAngle
SoSFEnum	precisionMode
SoImageVizEngineOutput < SoSFImageDataAdapter, SoImageDataAdapter * >	outImage

Detailed Description

SoRotateProcessing2d image filter

The SoRotateProcessing2d filter performs a rotation of an image by a user-defined angle $\theta$ and an origin $O$ . The new coordinates $(x', y')$ can be expressed as a function of the old coordinates $(x, y)$ :

$\left\{\begin{matrix}x'&=&(x-x_0)\cdot\cos\theta - (y-y_0)\cdot\sin\theta + x_0\\ y'&=&(x-x_0)\cdot\sin\theta + (y-y_0)\cdot\cos\theta + y_0\end{matrix}\right\}$

where $(x_0,y_0)$ are the coordinates of the center of the rotation. Or in a matrix notation:

$\begin{bmatrix}x'\\y'\end{bmatrix} = \begin{bmatrix}\cos\theta&-\sin\theta\\\sin\theta&\cos\theta\end{bmatrix} \cdot\begin{bmatrix}x-x_0\\y-y_0\end{bmatrix} + \begin{bmatrix}x_0\\y_0\end{bmatrix}$

Once again, destination pixels may be outside the image, and normally these pixels are ignored. Furthermore, the $(x, y)$ coordinates obtained are not always integers, even though an image is a discrete space. The figure below illustrates the grid of the resultant image, the point $M'$ being a pixel in this image. The points are generated from the rotation of the original image and do not fit on the grid.

Figure 1: Rotation

To calculate the intensity of each pixel $M'$ , two methods are possible:

Take the grey level of the nearest neighbor. For the above example would be given the intensity of
Take into account the four nearest neighbors and calculate the grey level based on the average of the four points, weighted by their distance, as in:
$g'=\frac{f_1\times g'_1+f_2\times g'_2+f_3\times g'_3+f_4\times g'_4}{f_1+f_2+f_3+f_4}$

If $d_j$ denotes the distance $M'_j M'$ , $f_i$ is defined as:

$f_i=\prod_{\substack{j=1\\j\neq i}}^{j=4}d_j$

e.g. $f_1=d_2d_3d_4$ .

The choice of the $f_i$ 's ensures that the interpolated value is equal to $g'_j$ if $M'$ matches $M'_j$ . This method gives better results but requires more computation time.

NOTE: In SoRotateProcessing2d filter, the image is perceived as a cylinder, where the information outside the image area wraps around and is placed in the blank part.