How can I count the number of leafs present in a manuscript bundle in an image using image processing?
Asked
Active
Viewed 57 times
0
-
can you manually do the counting? – Shai Jan 28 '16 at 10:46
-
1I think my comment was misunderstood: Looking at the attached image, I'm not sure the information is there. That is, even the ideal image processing algorithm that can do anything, will not be able to count the number of leafs simply because the resolution at which the image is taken is not enough. – Shai Jan 28 '16 at 11:51
-
Start with taking an image with MUCH MUCH higher resolution: in order to do the counting, you need that each leaf will be (much) larger than a pixel. – Shai Jan 28 '16 at 11:59
-
1Before you start thinking about doing a task like this, try to write a pseudo-code. If you can't explain with words _exactly_ what you want your code to do then you are very likely to fail. If you already have a concise idea for how you want to do this then please share it! For instance: 1, Find the angle of the image and rotate it so that each page is horizontally aligned. 2. Find a _very specific_ feature that differentiate the pages from the rest of the manuscript and crop the picture. 3. Convert it to black and white. 4.Count the number of black lines. But first of all, get a better image! – Stewie Griffin Jan 28 '16 at 12:04
-
I agree with @Shai. It looks like that there is no enough information on that image to count the amount of pages. – Ander Biguri Jan 28 '16 at 14:48
-
@StewieGriffin if you want to find the angle of the image and horizontally align it, you can find [this tutorial](http://stackoverflow.com/a/32228759/1714410) useful ;) – Shai Jan 28 '16 at 14:51
-
1That's an awesome answer @Shai. I remember it actually, you had my upvote 6 months ago! :-) – Stewie Griffin Jan 28 '16 at 15:36
-
1There are existing algorithms to do this, tho' mostly proprietary. For example, if the pages are reasonably well-maintained, a high-resolution scan will generate a vaguely sinusoidal output because the brightness varies from edge to "between-edge" . If you need an exact count, it's unlikely that any imaging system will work. If you just need, say 99% accuracy, it's possible. – Carl Witthoft Jan 28 '16 at 15:43