hough transform distance

i have finished making my cbir system but the result i get is not what is supposed to come out.

the query image i use to search for in the database the exact same image is not a part of the first 8 results .

i saw that once i stepped through my code. and saw that the distance i calculated for the query image was like 1787 or something and the distance calculated for the same image stored in the database came to be 2350 to about 3000 why is this the query image is original colour image as for the image stored in the database is a cropped image which i previously processed of doing edge detection and then saving that image to database. the way i calculated the distance was that i calculate hough transform of an image and square each entry in the resulting matrix and that is squared rooted and that is the distance.

i want to find a way to calculate the distance of an imagethat makes it unique so that it increases the chance of every different combination of a feature in an image is different.

Thanks

Answer 1

[Walter Roberson]

Goedelization would be one way. Radix or mixed-radix representation would be another.

You might want to use Bloom Filters to probe to see whether the image is in the database.

[Jack]

what if i was to use the y values of the peaks which i plot on the hough matrix to calculate distance shouldn't that work.