# tilted integral sum

Hello all! Is getting of tilted integral sum is necessarily for implementing Viola-Jones method? It seems to me that I saw some realizations which don't use tilted integral sum. This kind of integral image is some strange for me, I can't understand it...

edit retag close merge delete

Sort by ยป oldest newest most voted

No, the tilted integral image is not needed for the Viola Jones algorithm. In fact, the original Viola Jones Paper only uses upright (=untilted) Haar-like features.

In An Extended Set of Haar-Like Features for Rapid Object Detection, Rainer Lienhart , Jochen Maydt, ICIP 2002 the authors showed that extending the original feature set with those tilted Haar-like features improves performance, i.e. the false positive rate was reduced.

AFAIK the Haar-cascade detector in OpenCV uses both normal and tilted Haar-like features by default. I suppose other object classes benefit more from tilted features than faces.

more

( 2013-02-20 02:53:06 -0500 )edit

I have no clue.

( 2013-02-20 04:56:05 -0500 )edit

ok, thanks

( 2013-02-20 05:11:50 -0500 )edit

the viola-jones method is all about comparing rectangles, is the left one brighter than the right ?

with those integral images, the check is just a handful of array lookups and adds, without, you'd have to loop through the image region every time, to get it's intensity

more

This is not the answer for the question. It is just about integral images in general.

( 2013-02-19 16:34:04 -0500 )edit
1

well oh, i assumed the guy's problem was understanding integral images, thus a wish to sidestep them. my intent here was more to highlight the importance of the 'integral' thing for this.

( 2013-02-19 16:56:46 -0500 )edit

Yeah, maybe I say not very clear... I can't understand only titled integral image. Anyway thanks to the answer!

( 2013-02-20 00:29:43 -0500 )edit

Hi Bibliarij, i'm facing same problem like you, I can't understand tilted integral sum. Did you understand it now?

( 2013-07-24 21:50:23 -0500 )edit

Official site

GitHub

Wiki

Documentation