Go Premium for a chance to win a PS4. Enter to Win

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 165
  • Last Modified:

To fva,could you explain it to me in detail?

  hi,dear fva,in my previous question about how to compare the two id pictures to decide whether they are the same person or not,you answered me with these information:
"I agree, it's a complicated issue. On a simplified environment (same background, roughly same face size,
something like person pictures on ID cards) you can try something like this:

1. equalize the contrast and the brightness of the images
2. break the images in small squares. How small - you should experiment
3. Compare then the corresponding squares. If you have color info in the images, go for color comparision
first, then for the luminance component.
4. Offer a decision based on some treshold. Which one - you should experiment."
   

    Now,here is my question:1.why should we equalize the  contrast and the brithtness of the images firstly?Could you tell me?
                            2.why should we break the images in small squares?Any convincible reasons for it?
    Hope to hear from you soon.Thank you.

0
huaqiangsheng
Asked:
huaqiangsheng
2 Solutions
 
robert_marquardtCommented:
I do not understand much of picture comparision.

1. make the pictures match better.
This is done for any movie or the scenes would look awful.
How do you expect to match a picture where not a single pixel matches in color or brightness?
2. this makes it easier to find the relevant parts of the pictures. Comparision effort is quadratic.

You should get OpenCV library from Intel. Its free. It has a mailing list to discuss problems. It has many advanced functions for picture comparision.
0
 
fvaCommented:
Robert is right, but as being the addressee of the question I feel obliged to reply explicitely myself:

1. I haven't done this myself, but I was looking into image comparision issues lately and there is always the suggestion about equalizing. The reason is that if you have the very same image but with altered brightness or contrast, the two would seem different even to the human eye, let alone the dumb computer.
2. Breaking the images in small squares would let you to compare for similarity while keeping a reasonable level of tolerance. If you compare pixel-by-pixel, the slightest movement of the face as its global position between image 1 and image 2 will lead to a false "different" decision. If you extend to small squares and compare their average content, then those small movements will average out.

If you can enforce that the images are always aligned the same, I would suggest also that you assign different levels of significance to different squares on the image. The edge ones will probably contain the background anyway, and they are good candidates for reference points in equalizing the brightness and the contrast, while the ones "inside" the face area are better candidates for comparision, with higher significance on certain areas of the face.

F.
0
 
DragonSlayerCommented:
huaqiangsheng,
No comment has been added lately (823 days), so it's time to clean up this TA.
I will leave a recommendation in the Cleanup topic area for this question:

RECOMMENDATION: split points between robert_marquardt http:#6151564 and fva

Please leave any comments here within 7 days.

-- Please DO NOT accept this comment as an answer ! --

Thanks,

DragonSlayer
EE Cleanup Volunteer
0

Featured Post

What does it mean to be "Always On"?

Is your cloud always on? With an Always On cloud you won't have to worry about downtime for maintenance or software application code updates, ensuring that your bottom line isn't affected.

Tackle projects and never again get stuck behind a technical roadblock.
Join Now