Edge detection and cropping out region of higer intensity of Grayscale images

Hi all,

Please find attached image that I am working on. The region of interest that I intend to crop out (and to keep values of x & y )is shown to be the parabolic fluid profile of higher intensity than its background.

I have tried Edge of 'zero crossing' and 'canny', which are able to standout the ROI by slightly. I am however more interested in a total cropping of the ROI leaving all background as 0 in binary. I suppose my images are not like most of the examples used for edge where there is a clear distinction/ continuous distinction between edges, rather mine are dotted. I came across someone else did a median filter then a std deviation filter then only "canny" edge.

I have been able to use improfile to have a look at those intensities values, and the first and the last peak across the X -direction marks the boundary of the ROI.

I could it dot or outline the interface manually using data cursor, but it is definitely not ideal. I could it for just 1 frame manually, however not with a moving interface with each frame.

Please advice what others method should I try or use? Much appreciated for any inputs or assistance.

Colin

[Colin Ng]

Hi Mukul,

Thanks for the response.

I came across this error when running through your code.

Undefined function 'bwareafilt' for input arguments of type 'double'.

I have also tried other 'bwareafilt' stock example which is available online, to no avail. I have image processing toolbox ver 9.0 with Matlab R2014a installed.

Your code seems fine to me, I am sure I just needed to get the function sorted. Thanks, Colin

[Image Analyst]

bwareafilt() is a fairly new addition to the Image Processing Toolbox - you must have an old version. I am almost an expert in image processing (at least I'm still trying after 36 years of doing nothing else but image processing), but I can't do much with your stuff because your BMP image is corrupted (it says so when I click on it) and you forgot to attach your script.

Answer 1

[Mukul Rao]

Hi , I am no expert in image processing but here is what I came up with after playing around with some of the tools, I hope this helps! I believe the method I have used should be evident from the commands and code comments:

    %Read the image
    I = imread('1405ExpIP20895.bmp');

    %Convert to binary based on a threshold
    BW = im2bw(I, 0.1);

    %Improve connectivity
    fillBW = imfill(BW,'holes');

    %Select the largest object
    BWfilt = bwareafilt(fillBW,1,'largest'); 

    figure
    imshow(BWfilt)

    %Smoothen the boundaries
    BWfinal = imopen(BWfilt, strel('disk',9));
    figure
    imshow(BWfinal)

    %Index into original image
    I(~BWfinal) = 0;
    figure
    imshow(I)

[Colin Ng]

To update on my situation, I tried it on 2011b and 2014a, bwareafilt still can't be loaded. I have then tried it out with 2015a, and it is working fine.

I am trying to fine tune the process, as some of the images are not being cropped out appropriately. Thanks for your help, Mukul.

Colin

[Colin Ng]

Hi Mukul,

I have had some success with your code, thanks again. These 2 images attached show that some region could not be cropped out as per desired. It is to do with the imfill function, I believe. Maybe it needs some kind of manual intervention. I am looking into imfill.