Linear Filtering¶

The convolution functions use linear combinations of neighboring pixels to highlight some features in an image.

All the convolution functions can be destructive, meaning that the destination image overwrites the source image.
eVision offers 2 ways of doing convolutions:
- Use a predefined method with a predefined filter, and in some cases, a kernel size.
- Create your own EKernel ^.NET ^Python and use the function EasyImage::ConvolKernel ^.NET ^Python.

Predefined Filters¶

The available predefined filters are:

When you use your own kernel:

You can choose the width and height of the kernel.
- Use EKernel::SetKernelData ^.NET ^Python to fill the convolution coefficients.
eVision automatically normalizes the kernel coefficients so that their sum is 1.
- If you define a EKernel::SetGain ^.NET ^Python (multiplying all resulting pixels by that value) and an EKernel::SetOffset ^.NET ^Python (adding that value to all resulting pixels), they are applied after that normalization.
- Note that using any gain other than 1 can lead to a saturation in the resulting image and an overflow or an underflow during the internal calculations.
To rectify any value (to keep only the positive ones for instance), set an EKernelRectifier ^.NET ^Python.
The outside value is used for the border calculations (when the kernel needs values from out of the image on the border of the image to compute the result).
Another way to create a kernel is by giving the constructor an EKernelType ^.NET ^Python. This is quite similar to the way of doing a convolution with a predefined filter.