CS7321 Winter 1998

PS2 Solution by James Mossman

Two problems were encountered while working on PS2:

1) LP/HP filtering would "see" the boundary of the image as a transition region and blur/enhance accordingly. As a result, the PSD plot often contained information related to the orientation of the exterior boundary in addition to 'true' image orientation data.

2) The statement "log square modulus of the Fourier transform" seemed rather cryptic. After examining the data which resulted from the Fourier transform, a 'hack' PSD function was constructed which attempts to produce images which match those from the "Computer Vision" (chapter 2) handout.
 

Original Image	Operation (LP or HP)	Power Spectrum
	N/A
	low pass filter
	high pass filter
	N/A
	low pass filter
	high pass filter
	N/A
	low pass filter
	high pass filter
	N/A
	low pass filter
	high pass filter

• script.m: top-level script used to generate all images shown above (except for initial images)
• psd.m: function which generates the power spectrum of a given JPG. I had a lot of trouble trying to decipher the meaning of "log square modulus of the Fourier transform". The resulting images seem to roughly match the "Computer Vision" handout, so my 'hack' has to be fairly accurate.
• lp.m: function which applies a low-pass filter to a given JPG. The function also calls 'psd()' to calculate the corresponding power spectrum.
• hp.m: function which applies a high-pass filter to a given JPG. The function also calls 'psd()' to calculate the corresponding power spectrum.