Digital Signal Processing Reference
In-Depth Information
4.
SIMULATIONS
If we only need almost 30 dB picture quality for the reconstructed image
then we can use a directional difference value described above and we do not
need information almost DWT-details at all. That is, the image
reconstruction is performed through the DWT-approximation only. A higher
picture quality, however, when we need, DWT-detail information is needed.
Figures 14-6 and 14-7 are the reconstructed image with an average bit rate
0.15 bits/pel for the DWT-details. The SNRs of these are 34 dB (lenna) and
35 dB (Happa) and these total average bit rates are both 0.47 bits/pel. The
DWT-approximation is coded by using an ordinary JPEG method and its
average bit rate is 1.28 bits/pel. The picture qualities of these reconstructed
images are about the same as those of JPEG 2000 and look excellent for a
practical use.
5.
CONCLUSION
A new image compression method using DWT has been presented. In this
work DWT-details are estimated through its DWT-approximation. The
directional one-dimensional difference is introduced for this estimation.
Horizontal, vertical, and diagonal DWT-details have the different
corresponding one-dimensional differences respectively. In case of Haar
mother wavelet these directional one-dimensional difference is worthwhile.
We utilize our presented method only for level one-DWT of an image. It
looks interesting to use the method also for level two-DWT. It will be
considered as a future problem. In any case making IC and/or LSI with the
presented method is easy because of Haar wavelet and one dimensional
difference. Hence applying this method to a mobile phone is one of the best
utilities.
