[Image-SIG] PIL 'special' edge detection problem

[jcupitt at gmail.com]

Hi,

2008/9/16 Hans-Peter Jansen <hpj at urpla.net>:
> Now, I try to detect these folds. My 'solution' misses any sophisticated
> image processing algorithm, I simply compare one pixel with the one below,
> convert it to HSV (with ranges: 0..255, 0..100, 0..100), combine h and v
> deviations into a value, and try to find horizontal runs of patterns with a
> high enough value.

You can make this a lot quicker if you use image processing operations
rather than looping yourself for each pixel. I don't actually know the
PIL ones very well :-( but try something like:

* go to a signed monochrome image

You need to be able to handle negative values, and I don't think PIL
can do RGB int. Anyway, colour won't be much help here.

* subtract every pixel from the one below

You can do this with a convolution. Just use a mask like

  1
  -1

And each pixel will become the difference between itself and the pixel
below. Most of these values will be rather small, but if one of your
folds is present, you'll see sudden large spikes.

* take the absolute value

Now you don't need negatives!

* find the average value for each row

You'd normally do this with a projection operator (sum every row and
every column), I'm not sure if PIL has one of these. Shrinking the
image to a single column should do the same job.

* find peaks over a threshold

Now loop down the image and look for values over a certain threshold,
perhaps 50?

John