| Abstract |
We present two approaches to object detection:
- rstly, object detection using matched wavelets and,
- secondly, wavelet based technique for object detection in image
sequences.
From the lter-bank interpretation of wavelet transform the
wavelet should ”match” the signal of interest such that the output
of the matched lter bank is maximized. The wavelet transform
is a correlation transform and the correlation is maximized when
the spectrum of the wavelet (for corresponding value of dilation
and translation) is matched to the spectrum of signal of interest.
In the second approach we apply wavelet based technique to
detect object in the image sequences. We dene two hypotheses to
motion detection. To characterize motion we used some heuristics
criteria.
|
