In image processing, kernels have effects on two image properties:
        i) Image noise.
        ii) Spatial resolution.

Table 1372 lists various types of kernels which are commonly used in analysis of EM (electron microscopy) images and elemental maps.

Table 1372. Kernel types.

Kernel type	Remarks
Smooth Kernel	Has the effect of smoothing object boundaries, but as a result it reduces spatial resolution.
Sharpen kernel	Improves spatial resolution by enhancing object boundaries but at the cost of image noise so that it is used mainly to increase the brightness at edges and boundaries.
Laplacian kernel	Is defined as the (pseudo)inverse of the Laplacian. ( α> 0)
Sobel kernel	Is named after Irwin Sobel, and is less sensitive to noise but produces a rather poorly localized description of the edge feature.

In the image enhancement of image processing, one method is to enlarge the intensity differences within a region of interest (ROI) in the image. These gradient filters enhances edges, and thus the apparent sharpness of a detail is increased due to the increased contrast at the edges. The size of the ROI defines what size of objects are emphasized. Therefore, this filter is characterized by the kernel size, which is equal to the number of pixels in the ROI. Assuming a pixel size of 0.2 x 0.2 mm, the kernel size is 1 x 1 mm and then the objects of this size will be enhanced. The kernel size determines the size of enhanced structure: small ROI only enhance small objects, while large kernels only enhance large objects. On the other hand, the amount of enhancement is determined by the boost or enhancement factor.