Basic Image Filtering Operations
Image filtering involves the application of window operations that achieve useful effects, such as noise removal or image enhancement. This chapter deals particularly with what can be achieved with quite basic approaches, such as application of local mean, median, or mode filters to digital images. The focus is on gray-scale images, though some aspects of color processing are also covered.
• what can be achieved by low-pass filtering in the spatial frequency domain.
• how the same process can be carried out by convolution in the spatial domain.
• the problem of impulse noise and what can be achieved with a limiting filter.
• the value of median, mode, and rank order filters.
• how computational load can be reduced.
• what can be achieved by sharp-unsharp masking.
• the distinction between image enhancement and image restoration.
• the distortions produced by standard filters—mean, Gaussian, median, mode, and rank order filters.
This chapter delves into the properties of a variety of standard types of filters in order to show what they can achieve and what their limitations are. The edge shifts produced by most of these filters are small but predictable, and therefore correctable in principle. The exception is the rank order filter, for which the shifts can be large—but then this is the advantage of this type of filter and is at the core of mathematical morphology (see Chapter 8).