ABSTRACT

This chapter introduces the frequency domain approach to image filtering. It discusses the contributions of high and low spatial frequencies to the appearance of an image and this led to a generalized view of filtering in the frequency domain. The chapter describes implementation of the generalized view using the Fourier transform, and the reader performed frequency domain filtering using ImageJ. Shapes in images are made up of changes in gray level across the image, from dark to light to dark. To distinguish this method from filtering in the spatial domain, the process is called filtering in the frequency domain, or in frequency space. In other words, it performs frequency domain filtering using the same kernel as was used for convolution filtering. Frequency domain filtering is much faster than spatial convolution when the convolution kernel is larger than a few values.