Demonstration of a selection of image processing operations

David Young

Setup
Edge detection
Thresholding
Finding connected regions
Finding the sizes of different regions
Finding blobs in a certain size range (100-200 pixels)
Experimenting

Setup

This assumes that the Sussex vision library is in your Matlab path.

Read in an image, and display it.

Image = teachimage('vcr.bmp');
imshow(Image);

Edge detection

Experiment with the two kinds of edge detectors described in the lectures. We use the implementations from the Matlab Image Processing Toolbox, accessed via the edge function.

edges = edge(Image, 'log', [], 1);     % Marr-type, sigma = 1
figure; imshow(edges); title('Marr edges, sigma = 1');
edges = edge(Image, 'log', [], 2);     % Marr-type, sigma = 2
figure; imshow(edges); title('Marr edges, sigma = 2');

edges = edge(Image, 'canny', [], 1);   % Canny, sigma = 1
figure; imshow(edges); title('Canny edges, sigma = 1');
edges = edge(Image, 'canny', [], 2);   % Canny, sigma = 2
figure; imshow(edges); title('Canny edges, sigma = 2');

Thresholding

Matlab provides a thresholding function, but the operation is most simply written (for ordinary arrays of type double) using the arithmetic "greater than" operator.

thresh = 0.5;

bin = Image > thresh;      % regions brighter than thresh
figure; imshow(bin); title(['White is where original image > ' num2str(thresh)]);

bin = Image < thresh;      % regions darker than thresh
figure; imshow(bin); title(['White is where original image < ' num2str(thresh)]);

Finding connected regions

We use the toolbox function, with 8-connectivity.

labelled = bwlabel(bin, 8);
figure; imshow(labelled, []);
colormap(jet);    % different regions different colours

Finding the sizes of different regions

We find areas as an example, but lots of other statistics may be measured - see help for regionprops. The second line of code extracts the sizes from the structure array to an ordinary array. For more information, see the Matlab documentation on structure arrays.

The number of blobs is printed.

blobstructures = regionprops(labelled, 'Area');
blobsizes = [blobstructures.Area];
no_blobs = length(blobsizes)

no_blobs =
   135

Finding blobs in a certain size range (100-200 pixels)

We can easily pick out the blobs in a size range of interest. find is a useful function that returns the indices of non-zero elements of an array. In this case, the indices returned are the same as the labels of the blobs in the size range.

We then select just the blobs with these labels, using ismember to test each pixel of the labelled array to see if it belongs to one of the blobs of interest.

minBlobSize = 100;
maxBlobSize = 200;
special_blobs = find(blobsizes >= minBlobSize & blobsizes <= maxBlobSize)
some_blobs = ismember(labelled, special_blobs);
imshow(some_blobs);

special_blobs =
  Columns 1 through 7
     7    11    64    70    71    75    76
  Columns 8 through 10
    82    83    92

Experimenting

You can download this document and then extract the original M-file with Matlab's grabcode function. You can then edit it for experimentation. (Functions from the Sussex vision library are only available to Sussex students and staff.)