Image analysis and feature extraction

Dear all I am working on a project dealing with images. I want to extract the features of each image then combine all features into one table( database).

Note: I did one project before but each image has some tissues (blobs) so I dealt with tissues as a record not the image. But in this case I need to deal with the image as a record.

Thank you

Answer 1

[Image Analyst]

I don't understand what you mean when you're talking about doing image analysis as an image versus a record. This makes absolutely no sense to me: "I dealt with tissues as a record not the image. But in this case I need to deal with the image as a record." You want "as a record" in both cases but what's the difference between "tissues" in the first case and "image" in the second case?

If you want to know how to measure some things in an image, see my image segmentation tutorial: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

[hanem ellethy]

Sorry for my way in explaining. Before, I segmented the image into some tissues (pieces) then extract the features using (regionprops) so I got the features for each tissue in the image. Now I don't want to segment the image, I need to extract features without segmentation.

[Image Analyst]

To do that, just have one gigantic blob that is basically the whole image. Then anything you measure from separate blobs before will now be measured as if the image were one giant rectangular blob.

% Make mask the entire image.
binaryImage = true(size(grayImage));
% Now make measurements
measurements = regionprops(binaryImage, 'All');

[hanem ellethy]

but this code give me a white image (all ones) that has the same size as grayImage

[hanem ellethy]

okay let me explain more. the attached images has two points and i want to extract some features in the images related to those two points but as a one blob as you mentioned above.

[Image Analyst]

Well you said "I don't want to segment the image, I need to extract features without segmentation" so if you don't segment it - meaning you don't break it up into separate regions - then all you have is one giant rectangular region that is the whole image.

The only feature a single pixel would have is its gray level. What other features do you think a single point (single pixel) should have?

[hanem ellethy]

okay, please advice how can I extract features for the previous image related to the two points in the image?

when I tried segmentation and regionprops I got two blobs and (2x1)structure array

when i tried to use something like this

regions = detectMSERFeatures(I);

I got a message means that it can't be done for unit 8 image

L7=imread('fig7.jpg');

L7=rgb2gray(L7);

LL7=imcomplement(L7);

figure(1);imshow(L7);title('original image');

h = imrect(gca, [169 5 989 788]);

maskImage = h.createMask();

% Mask the image with the rect.

LL7(~maskImage) = 0;

figure(2),imshow(LL7);title('ROI after applying mask');

binaryImage7 = LL7 > 100;

binaryImage7 = bwareaopen(binaryImage7,30);% Get rid of small specks of noise

binaryImage7 = imfill(binaryImage7, 'holes');

tt7 = regionprops(binaryImage7, 'all');

please help may be I can not explain good

[Image Analyst]

Sorry, I don't know what to tell you. You're now segmenting the image. I thought you did not want to do that. So you can either segment and measure the image, like you're doing here. Or you can not segment the image and just measure the whole image as one giant rectangular blob. Or you can look at just two points like you mentioned before and the only thing you can get from that is the gray levels or colors at those locations and the distance between the two points.

Regarding your error message, you can't convert to gray scale with rgb2gray if your image is already gray scale. So check and only call it if it's gray scale

if ndims(L7) == 3
    L7=rgb2gray(L7);
end