Hi,
Just after one level wavelet decomposition I can get edge information, which can be used as a feature. So why do I need to make two, three... more level wavelet decompositions? Moreover, due to more decomposition level is a reason to loss edge information. What is the benefit of multi-level wavelet decomposition? Please explain.
Multilevel wavelet decomposition provides you the information about frequency components present and enhance the information about the signal for further processing.
Thanks Kishan Rao Kalitkar sir. Could you explain it with an example?
Do I need to go lower level to capture edge features? When do I go for higher level? What is image detail?
Thanks Aparna Murthy madam for your help.
Does approximate image contain approximate coefficients?
I also suppose that first level of wavelet decomposition is sufficient for detection of thin effects. While multi-level decomposition gives you the means for best compressing of data.
In order for you to understand why you need more levels of decomposition I suggest you to play with zeroing individual subbands and after that to do inverse transform to see the difference. Particularly, you can see that for sufficiently smooth images/signals the subbands of the first level of decomposition don't contain edges at all. Another example is noisy image/signal, where the first level subbands contain edges, but you hardly recognize them because of noise, while the subsequent levels will contain more and more useful information and less noise. Generally several levels of decomposition allows to analyze and process image/signal features of different scales, which can be critical in some applications.
Besides, it is that very reason why they usually advise to use Interpolating Cubic Spline Wavelet Packet on Arbitrary Partitions
Article
Full-text available
Jianzhong Wang
1 Recommendation32 Reads11 Citations
Download
instead of semi- and orthogonal wavelets.
- Badges
- Science topic
- Similar topics
- Comment