MimConvolve

This function performs a general convolution operation on the source buffer using the specified filter, storing results in the specified destination buffer. This function supports two types of filters: Finite Impulse Response (FIR) filters and Infinite Impulse Response (IIR) filters. FIR filters operate on a finite neighborhood, while IIR filters take into account all values in an image. You can specify either a predefined FIR filter or a custom FIR or IIR filter.

Note that predefined FIR filters are the fastest type, and are recommended as the first to try if you are not sure which filter will best suit your needs. When using custom filters, relatively small FIR filters are faster and allow control over the kernel contents. If the FIR kernel that you require becomes too large, try an IIR filter. If speed is not an issue, try the adaptive filters available using MimFilterAdaptive(), which remove noise in an image while preserving edges as much as possible.

This function provides a number of predefined FIR filters. For predefined FIR filters, the overscan is set to transparent overscan pixel values, saturation is enabled for results, and the kernel's default center pixel is set as the top-left pixel of the central elements in a neighborhood. When using predefined FIR filters, you cannot change operation control settings except for the overscan.

This function also accepts a custom FIR or IIR filter with which you can customize operation control settings. To define a custom FIR filter, you must allocate a kernel buffer using MbufAlloc1d() or MbufAlloc2d(). Note that the kernel buffer size can be constrained by the available resources. To define a custom IIR filter, you must allocate a linear IIR filter image processing context, using MimAlloc() with M_LINEAR_FILTER_IIR_CONTEXT.

To define a custom FIR filter, you must load the kernel buffer with values, using MbufPut(). When using a custom FIR filter, you can change operation control settings using MbufControl(). To define a custom IIR filter, use MimControl() with a linear IIR filter image processing context and specify appropriate M_FILTER_TYPE, M_FILTER_OPERATION, M_FILTER_RESPONSE_TYPE, M_FILTER_SMOOTHNESS_TYPE, and M_FILTER_SMOOTHNESS operation control type settings.

For predefined and custom FIR filters, when using 32-bit integer buffers and saturation is disabled, the accumulator buffer is a 32-bit internal buffer (MIL_INT); whereas, when saturation is enabled, the accumulator buffer is a 64-bit floating-point internal buffer (double). For custom IIR filters, operations are always performed using floating-point precision.

For custom FIR filters, this function will internally separate a large kernel into two 1-dimensional kernels, if possible, to increase the speed of the convolution operation. For more information on separating kernels, see the Finite Impulse Response (FIR) filters subsection of the Custom spatial filters section of Chapter 4: Advanced image processing.

Custom FIR filters containing a single kernel value (such as uniform kernels or average kernels) are dealt with by a specifically targeted algorithmic optimization.

If the source and destination are multi-band buffers, the same filter is applied to every band of the source buffer.