Cropping greyscale image based on pixel

Hi all,

I intend to crop images based on its pixel value or intensity. Here are my images below, labeled as image1, image2... image 7.

All these images are from the taken from the object but replaced with different optical density filters. I performed a simple imcrop() rectangle and subplotted to obtain the following.

For example if I take image3, I am mostly interested in the center and would like to crop out that weird eclipse area.

I would like to somehow be able to define a range to crop and keep the cropped area. For each image, I will need to define a new range, depending on the section I want to keep or discard. Can I do something like, if the pixelvalue < 50, then make it transparent?

Later on, I would like to paste these cropped areas on top of each other.

Thanks, Victor

Answer 1

[Peter O]

The quick-n-dirty way that comes to mind is to just use a bitmask: 0 to toss a pixel's value and 1 to keep it. It's a little memory intensive but it won't care about the shape of the ROI.

For a given filtered image,

    %Returns a logical matrix equal in size to your image, doing a per-pixel comparison.
    Mask1 = (Im1 > 50);

Then to combine, just add everyone together with the bitmasks, performing a bitwise mult to only keep the pixels you want. The other pixels will be zero. So not exactly transparent, but since you're working in grayscale I suspect you don't have (or need) an alpha channel.

ImAll = Im1.*Mask1 + Im2.*Mask2 + ...

Obviously you could use N-D arrays instead of discrete variables and matrices and a half dozen other tricks if you wanted, depending on your end application. Is that sufficient? Or too coarse a solution?

[victor law]

Thank you for your quick reply,

I have had an attempt at your given code and it is cropping the desired region. As you mentioned above, pixels excluded become 0, but the rest becomes all white.

The next step is how can I keep the data values within my range?

How do I retain the grey values in the cropped region rather than displaying it all white?

Many thanks,

[victor law]

The code seems pretty small, but all I have done is,

 c06 = double(b06) %Original Image1
c15 = double(b15) %Original Image2
Mask1 = (c06 > 200) 
CroppedImage1 = Mask1.*c06 %Separating image based on pixel value for Image1
Mask2 = (c15 > 100);
CroppedImage2 = Mask2.*c15 %%Separating image based on pixel value for Image2

The rest is just a subplot comparison between original and cropped image.

[victor law]

Nevermind, the solution worked after playing around with it.

Thank you very much Peter!

Answer 2

[Image Analyst]

I'm not sure of the exact algorithms they use in HDR reconstruction. A simplistic view would be to extract non-saturated pixels and divide them by the transmissivity, then average or scale any pixels that show up in multiple images, as I said the prior time you asked this.