Hi Bazik

I remeber reading an article from a university that were developing a image recognition system based on the REM (Rapid eyes movement).

The image was scanned diagonally just like the way our eyes do when looking at something.

The eyes seems to scan an image beginning of the center. Several images was put together and the program identified the one the user choose before.

So, when the program is runninf...he "traces" on the image a path simulation your eyes direction.

Then he make the same with the other files and compare the generated path.

The algo seems to work with brigthness and contrast of the images, since it work on black and white images only, i guess (grayed i mean).

I?m not at home, but i can look if i still have the link and send to you when i find.

Also, yesterday I have found a program that may be adapted to your needs. It converts an image in html file.

For example, you have a coloured bitmap of a face.....and when running the program it convertes the image to a serie of 0 1 0 1 0 1 0 ..etc.....forming the image on this way.

Since the result is in html, the 1 1 0 1 ......are couloured accordly to the image colors. The result is very good.

Something similar can be implemented to it.

After scanning the image, he will transform all to binary numbers (1 0 0 etc..), and place some specific data to identify the colors.

For black and white images, the recognitions can be make after counting the number of 1 or 0 on a line.

Scanning the hole picture will give you the amoun of 1 / 0 with each line.

This can be the beginning of an digital ID.

The development can be done as the same way as Flair (File language identification system).

For example....let?s say that on image 1 you have the face of a person with a black background. (To make easier to have the "Digital ID" of the face)...

So...The ID can be made in steps.....If the face is looking at the same position...(Identifying different faces, for example)

To get accuracy of the Digital ID. You can make several different kinds of combinations of brigthenss and contrast to check the variance of the 1 0.

So, for example...if the digital ID (of the 1st scanned line) using brightness 35% and contrtast 28% is:

1 0 1 1 1 1 1 1 0 0 0 1 0 1 0 1 1 0 0 1

Compare with another variances . Let?s say that at brightness 49% and contrast 27%, the ID is:

0 1 0 0 1 1 1 1 0 1 0 1 0 1 0 1 1 0 0 0

You just compare line 1 to line 2 and check if the values match at a certain position:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 0 1 1 1 1 1 1 0 0 0 1 0 1 0 1 1 0 0 1
0 1 0 0 1 1 1 1 0 1 0 1 0 1 0 1 1 0 0 0

X X X X 1 1 1 1 0 X 0 1 0 1 0 1 1 0 0 X

After several combinations you can see the amount of matched values and build a ID of each line.

Let?s say that at line 1 after 100 combinations, you have a percentage of 57 % of matched 1 and 38% of matched 0

This can be the bases of the signature id....

For example.....a sign of "X" will bypass the recognition value the rest is only matching at each position.
Ther ID for line 1 can be a combination of position and value

The result for line 1 can be:

Position(In hexa) + Value

51 61 71 81 90 B0 D0 etc etc etc....

This is just an idea, making an approach to the Flair system.

Best Regards,

Beyond2000!
Posted on 2003-03-24 11:33:46 by Beyond2000!
Beyond2000,

Good to see you back again.. you seemed to have fallen off the map...

:NaN:
Posted on 2003-03-24 23:08:19 by NaN
Hi Nan

I'm envolved on a huge project.

It's for the Spasm disassembler (and for the assembler, too).....is a very nice work Ren? is doing. and I'm colecting information of all Apis ever made before....about 8000 functions., 55000 equates, 1400 structures etc.

I come to the board from time to time, but last months a have been so busy.....and the work on the project is kinda slow, since it is a lot of data to colect and rebuild

I've got some volunteers to help me, but now only one nice person is helping with it....

I hope we may finish at least the structures in 02 weeks or less.

Take a look at it Nan

it's really interesting

Best Regards,

Beyond2000!
Posted on 2003-03-25 00:47:10 by Beyond2000!
hey!

bazik
i have a qestion . did you manage to get the regonize work well? can you publish the source?

thanks