Computes 2D convolution of canvas' image and given filter kernel.

public procedure ApplyConvolution(Kernel: PLongInt; 
                                  KernelSize: LongInt; 
                                  Divisor: LongInt; 
                                  Bias: Single = 0.0; 
                                  ClampChannels: Boolean = True); virtual;

Kernel is in row format and KernelSize must be odd number >= 3. Divisor is normalizing value based on Kernel (usually sum of all kernel's cells). The Bias number shifts each color value by a fixed amount (color values are usually in range [0, 1] during processing). If ClampChannels is True all output color values are clamped to [0, 1]. You can use predefined filter kernels or define your own.