Need help in pixel count!

Hi everybody! I'm struggling to get this one solved.

I have a binary image(640x480) full of 'Salt and Pepper' noise, and I need to count the pixels that are inside a imaginary square with dimensions 100x100, rotated 30º. The upper left pixel of the square (corner) has to be lying in the pixel (200,100) of the image.

I think the best way to solve this is to create a mask with the square rotated and then apply it to the image.

What would be the best way to get it done? I dont get how I guarantee that the upper left corner of the square would be in the pixel (200,300). I need to get a new image with the same size of the original with the square in the right position. Do you have any ideas to solve this?

Here's an image to show you what I mean. http://dl.dropbox.com/u/5272012/Untitled.png

Answer 1

[Sean de Wolski]

Do you want the upper corner at 200,300 or the lefter corner at 200,300. Looking at the image I can't tell.

Either way, you could use the second output from max(bw(:)) on your mask image and sub2ind() to find row/col coordinates. This will be the index of the first maximizer it finds, guaranteed to be a corner and guaranteed to be 1 in an binary image. Whether it's the top corner or the left corner will depend on whether you transpose the mask first.

[Walter Roberson]

"lefter" ?

[Sean de Wolski]

you got the gist right? ;)

I guess I should apply the Maine-accent

that vertex is leftah than the uppah one!

Answer 2

[Rui Trovisco]

Solved! Thanks!

[Image Analyst]

I'd guess it was via poly2mask(), thresholding, regionprops, then doing hist([measurements.Area], nBins). Once you have the corner coords, it's basically 4 key lines of code (more if you add comments, display images, print out results, etc.)