Page 422 - Introduction to Information Optics

7.7. Pattern Recognition with Photorefracthe Optics

P S U R M

X. = 670.0 nm

X = 670.1 nm
u

X» 670.3 nm
M

X = 670.4 nm

Fig. 7.44. Output correlation peak intensities show shift invarianee and the spectra! content,

reconstructed holographic images obtained with the specific spectral line are
shown in Fig. 7.43b. To test the shift invarianee of the PR filter, the five input
letters are inserted at the input plane at the same time. Notice that these letters
have been shifted with respect to the recorded PR filter. The corresponding
correlation peak intensities for each spectral line are detected by a CCD
camera. By properly thresholding these detected peak intensities, a set of
correlation peaks corresponding to the recorded spectral contents is obtained,
as depicted in Fig. 7.44. From these results, we see that the PR-based RC can
indeed preserve the shift invarianee by using a wavelength-multiplexed reflec-
tion-type PR filter.

7.7.3. WAVELET MATCHED FILTERING

Wavelet transform (WT) analysis is a viable alternative to Fourier trans-
forms for optical pattern recognition. The WT has been utilized for multi-
resolution image analysis and in optical implementations. The 2-D WT of a
signal is given by

a y, b x, h y)
(7,60)

y)h* h(x, y)clxdy,