Digital Signal Processing Reference
In-Depth Information
Frequency
content
(a)
(b)
Frequency
content
(c)
(d)
FIGURE 14.36
Magnitude spectrum plots for the noisy image and the noise-filtered image: (a) the noisy image; (b) magnitude
spectrum of the noisy image; (c) noise-filtered image; (d) magnitude spectrum of the noise-filtered image.
see noise spectral components, the spectral magnitude is further multiplied by a factor of 100. Once the spectral
value is larger than 255, it is clipped to 255. The resultant spectrum is displayed in
Figure 14.36
(b), where we can
see that noise occupies the entirety of the image.
To enhance the image, we apply a Gaussian lowpass filter. The enhanced image is shown in
Figure 14.36
(c), in
which the enhancement is easily observed.
Figure 14.36
(
d) displays the spectra for the enhanced image with the
same scaling process described just above. As we can see, the noise is significantly reduced compared with
Figure 14.36
(
b).
14.7
IMAGE COMPRESSION BY DISCRETE COSINE TRANSFORM
Image compression is a must in our modern media systems, such as digital still and video cameras and
computer systems. The purpose of compression is to reduce information storage or transmission
bandwidth without losing image quality or at least without losing it significantly. Image compression
can be classified as lossless compression or lossy compression. Here we focus on lossy compression
using the discrete cosine transform (DCT).
The DCT is a core compression technology used in the industry standards JPEG (Joint Photo-
graphic Experts Group) for still-image compression and MPEG (Motion Picture Experts Group) for
Search WWH ::
Custom Search