<< Chapter < Page | Chapter >> Page > |
Probably the simplest useful energy compression process is the Haar transform. In 1-dimension, this transforms a 2-elementvector into using:
Note that is an orthonormal matrix because its rows are orthogonal to each other(their dot products are zero) and they are normalised to unit magnitude. Therefore . (In this case is symmetric so .) Hence we may recover from using:
If then These operations correspond to the following filtering processes:
It is clear from that most of the energy is contained in the top left (Lo-Lo) subimageand the least energy is in the lower right (Hi-Hi) subimage. Note how the top right (Hi-Lo) subimage contains thenear-vertical edges and the lower left (Lo-Hi) subimage contains the near-horizontal edges.
The energies of the subimages and their percentages of the total are:
Lo-Lo | Hi-Lo | Lo-Hi | Hi-Hi |
---|---|---|---|
96.5% | 2.2% | 0.9% | 0.4% |
Total energy in and = .
We see that a significant compression of energy into the Lo-Lo subimage has been achieved. However the energy measurements donot tell us directly how much data compression this gives.
A much more useful measure than energy is the entropy of the subimages after a given amount of quantisation. This gives the minimum number of bitsper pel needed to represent the quantised data for each subimage, to a given accuracy, assuming that we use an idealentropy code. By comparing the total entropy of the 4 subimages with that of the original image, we can estimate the compressionthat one level of the Haar transform can provide.
Notification Switch
Would you like to follow the 'Image coding' conversation and receive update notifications?