Developer Reference for Intel® Integrated Performance Primitives
The fixed filter functions perform linear filtering of a source image using one of the predefined convolution kernels.
The supported fixed filters and their respective kernel sizes are listed in the following table:
| Fixed Filter Type | Kernel Size |
|---|---|
| Horizontal Prewitt operator | 3x3 |
| Vertical Prewitt operator | 3x3 |
| Horizontal Scharr operator | 3x3 |
| Vertical Scharr operator | 3x3 |
| Horizontal Sobel operator | 3x3 or 5x5 |
| Vertical Sobel operator | 3x3 or 5x5 |
| Second derivative horizontal Sobel operator | 3x3 or 5x5 |
| Second derivative vertical Sobel operator | 3x3 or 5x5 |
| Second cross derivative Sobel operator | 3x3 or 5x5 |
| Horizontal Roberts operator | 3x3 |
| Vertical Roberts operator | 3x3 |
| Laplacian highpass filter | 3x3 or 5x5 |
| Gaussian lowpass filter | 3x3 or 5x5 |
| Highpass filter | 3x3 or 5x5 |
| Lowpass filter | 3x3 or 5x5 |
| Sharpening filter | 3x3 |
Using fixed filter functions with predefined kernels is more efficient as it eliminates the need to create the convolution kernel in your application program.
The anchor cell is the center cell of the kernel for all fixed filters.