Image Enhancement Techniques

This chapter opens with the description of techniques to improve the visual qualities of the image, based on point, local, and global operators. Algorithms operating in the spatial domain and in the frequency domain are shown, highlighting with examples the significant differences between the described domains and algorithms, also from the point of view of the computational load. For the analysis of the spatial frequencies (horizontal, vertical, oblique) contained in the image, a very effective method is to perform a spatial frequency transform. This transform converts the image information from the spatial domain of the gray levels to the frequency domain (expressed in terms of magnitude and phase). The most widespread is the Fourier transform. Spatial filters are implemented through the spatial convolution process that processes the pixel value based on nearby pixel values. After defining the theory and implementation aspects of a linear operator, based essentially on the convolution process in the spatial domain and in the frequency domain, we illustrate many local smoothing operators for gray and color images. These operators are also intended to eliminate or attenuate the additive noise present in the image. This is achieved through local linear and nonlinear operators that essentially try to smooth out the irregularities in the image without altering the significant structures of the image itself. Linear smoothing filters can be defined in the spatial domain or in the frequency domain. For the spatial filters, in the convolution mask, the weights that characterize the particularity of the filter are appropriately defined. For filters in the frequency domain, similar effects on the image are obtained by removing high frequency components. In the description of the various effects provided by the filters, a frequency analysis of the spatial filter will often be used, and vice versa.