Here is my HW #6. This script uses only MATLAB internal functions. Please 
have objects.jpg in the path or in the working directory.

The script's outputs of note are Yt, the transformed Fourier analyses, Yts, 
the standard deviation, Ytm, the mean of values (to compute Yts), FSm, the 
mean of the Fourier Spectrum, Ytd, the Euclidean distance between the first 
10 frequencies of the first spectrum and ith specturm, and indx1, indx2, and 
indx3. indx1 is an index of the sorted list of Ytd, with each image compared 
to the first. Note that image 3 (the small gear) and image 14 (the wheel), 
are the highest sorted in this list -- both have roughly the same shape as 
the big gear. indx2 is an index of the sorted list of the least mean squared 
differences between the standard deviation of the high frequencies of image 
1 and the other images. Note that the similarity between the top items and 
the big gear are low. This is because there was a large disrepancy between 
the standard deviation of 1 and 3. More telling is indx3, an index of the 
sorted list of standard deviations. Images 3 and 1 top the list, and they 
are the only two with a std. above 0.1 which tells a lot more of their 
similarity than the lms.

Thus, with this method, it is possible to extract the gears -- those items 
with high std. dev. and a particular low band analysis (namely, a high 
amplitude primary frequency and low amplitude elsewhere in the low band).

Thomas Young