It's quiet here isn't it...

[Chris Greening]

Anybody there?

[Pradeepta Dash]

Hello Chris,
Thanks for building the library. I am having trouble extracting
connected regions from a sample image (in fact any image). I seem to
be getting only one object in the regions array with the bounding box
equals to the size of the image.
Is there any pre-processing needed (apart from converting to
greyscale)?

Could you provide a code sample, please?
(Presumably, sudoku grab is using the same code base to detect the
square region of the sudoku)

--pradeepta

On Mar 26, 4:49 pm, Chris Greening <cmgreen...@googlemail.com> wrote:
> Anybody there?

[Chris Greening]

Hi pradeepta,

I'll be updating the connected region extraction with a more efficient
version this week. I'll add sample code at the same time.

I think the step you are missing is that you need to threshold the
image so it only contains pixels that are 0 (background) and 255
(foreground).

For sudoku grab I threahold the image and then scan through looking
for the connected region with the most number if pixels. I assume that
this must be the puzzle.

Cheers
Chris



Sent from my iPhone