{"id":2614,"date":"2017-06-20T07:00:27","date_gmt":"2017-06-20T11:00:27","guid":{"rendered":"https:\/\/blogs.mathworks.com\/steve\/?p=2614"},"modified":"2019-11-01T17:12:50","modified_gmt":"2019-11-01T21:12:50","slug":"customizing-regionprops-with-your-own-measurements","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/steve\/2017\/06\/20\/customizing-regionprops-with-your-own-measurements\/","title":{"rendered":"Customizing REGIONPROPS With Your Own Measurements"},"content":{"rendered":"

I saw a presentation last month that mentioned a user request to have the ability to customize regionprops<\/tt>. That is, a user wanted to be able to add their own measurement to regionprops<\/tt>.<\/p>

Today, I'll show you how to do this yourself.<\/p>

First, here's a brief recap on what regionprops<\/tt> does. The function computes measurements of image regions. Some of these measurements are based purely on a region's shape, while others incorporate pixel values within the regions. Here's an example using the coins.png sample image.<\/p>

I = imread('coins.png'<\/span>);\r\nimshow(I)\r\n<\/pre>\"\" 
Let's convert this image to binary, using adaptive thresholding, filling holes, and removing small \"noise\" pixels.<\/p>
bw = imbinarize(I,'adaptive'<\/span>);\r\nbw = imfill(bw,'holes'<\/span>);\r\nbw = bwareafilt(bw,[100 Inf]);\r\nimshow(bw)\r\n<\/pre>\"\" 
You can count the \"blobs\" (object) yourself; there are 10 of them.<\/p>
The simplest regionprops<\/tt> call, regionprops(bw)<\/tt> computes the Area<\/tt>, Centroid<\/tt>, and BoundingBox<\/tt> for each object.<\/p>
s = regionprops(bw)\r\n<\/pre>
\r\ns = \r\n\r\n  10×1 struct array with fields:\r\n\r\n    Area\r\n    Centroid\r\n    BoundingBox\r\n\r\n<\/pre>
But I don't think this is the best way to call regionprops<\/tt> anymore. You can now tell regionprops<\/tt> to return the results as a table.<\/p>
t = regionprops('table'<\/span>,bw)\r\n<\/pre>
\r\nt =\r\n\r\n  10×3 table\r\n\r\n    Area        Centroid        BoundingBox \r\n    ____    ________________    ____________\r\n\r\n    2635    37.133    106.85    [1x4 double]\r\n    1846    56.131    49.693    [1x4 double]\r\n    2672    96.199    146.05    [1x4 double]\r\n    1839    109.97    84.848    [1x4 double]\r\n    2744    120.37    208.73    [1x4 double]\r\n    2520    148.57    34.404    [1x4 double]\r\n    2589    174.83    120.01    [1x4 double]\r\n    2518    216.81    70.649    [1x4 double]\r\n    1857    236.03    173.36    [1x4 double]\r\n    1829    265.96    102.64    [1x4 double]\r\n\r\n<\/pre>
The table form is a lot more convenient for many tasks. For today's topic, one especially nice thing thing about tables is how easy it is to add your own variables to the table.<\/p>
To illustrate, let's add a measurement that I've seen called Roundness<\/tt>. One definition for roundness is:<\/p>
$R = \\frac{4A}{\\pi L^2}$<\/p>
where $A$ is the object area and $L$ is the major axis length of the best-fit ellipse for the object. Here's how to compute roundness and add it directly to the measurements returned by regionprops<\/tt>.<\/p>
First, note that both Area<\/tt> and MajorAxisLength<\/tt> are supported by regionprops<\/tt>, so let's start with those.<\/p>
t = regionprops('table'<\/span>,bw,'Area'<\/span>,'MajorAxisLength'<\/span>)\r\n<\/pre>
\r\nt =\r\n\r\n  10×2 table\r\n\r\n    Area    MajorAxisLength\r\n    ____    _______________\r\n\r\n    2635     60.08         \r\n    1846    50.178         \r\n    2672    59.792         \r\n    1839    49.674         \r\n    2744    60.374         \r\n    2520     58.08         \r\n    2589    58.676         \r\n    2518    58.162         \r\n    1857     49.77         \r\n    1829    49.564         \r\n\r\n<\/pre>
You access table variables using dot notation, like t.Area<\/tt>. Similarly, you can create a new table variable using dot notation and assignment, like t.MyVariable = ...<\/tt>. So adding Roundness<\/tt> to the table returned by regionprops<\/tt> is this simple.<\/p>
t.Roundness = 4 * t.Area .\/ (pi * t.MajorAxisLength.^2)\r\n<\/pre>
\r\nt =\r\n\r\n  10×3 table\r\n\r\n    Area    MajorAxisLength    Roundness\r\n    ____    _______________    _________\r\n\r\n    2635     60.08             0.92945  \r\n    1846    50.178             0.93352  \r\n    2672    59.792              0.9516  \r\n    1839    49.674             0.94893  \r\n    2744    60.374              0.9585  \r\n    2520     58.08             0.95118  \r\n    2589    58.676             0.95745  \r\n    2518    58.162             0.94772  \r\n    1857     49.77             0.95453  \r\n    1829    49.564             0.94798  \r\n\r\n<\/pre>
Let's try this computation with an image containing objects that are not quite as round.<\/p>
I2 = imread('rice.png'<\/span>);\r\nimshow(I2)\r\n<\/pre>\"\" 
bw2 = imbinarize(I2,'adaptive'<\/span>);\r\nbw2 = imfill(bw2,'holes'<\/span>);\r\nbw2 = bwareafilt(bw2,[100 Inf]);\r\nimshow(bw2)\r\n<\/pre>\"\" 
t2 = regionprops('table'<\/span>,bw2,'Area'<\/span>,'MajorAxisLength'<\/span>);\r\nt2.Roundness = 4 * t2.Area .\/ (pi * t2.MajorAxisLength.^2);\r\nhead(t2)\r\n<\/pre>
\r\nans =\r\n\r\n  8×3 table\r\n\r\n    Area    MajorAxisLength    Roundness\r\n    ____    _______________    _________\r\n\r\n    138     23.594             0.31562  \r\n    120     18.152              0.4637  \r\n    169     28.123             0.27207  \r\n    157     23.793              0.3531  \r\n    284     43.757             0.18885  \r\n    200     26.259             0.36929  \r\n    141     21.647             0.38311  \r\n    177     29.087             0.26636  \r\n\r\n<\/pre>
I'm a big fan of the (relatively) new histogram<\/tt><\/a> function in MATLAB, so let's use it to compare our roundness numbers. I will follow the advice given in the histogram<\/tt> reference page<\/a> for normalizing multiple histograms so that they can be more easily compared. I'll set the y-axis limits to [0 1], which is appropriate for probability normalization, and I'll set the x-axis limits to [0 1], which is the range for Roundness<\/tt>.<\/p>
h1 = histogram(t.Roundness);\r\nhold on<\/span>\r\nh2 = histogram(t2.Roundness);\r\nhold off<\/span>\r\nh1.Normalization = 'probability'<\/span>;\r\nh2.Normalization = 'probability'<\/span>;\r\nh1.BinWidth = 0.02;\r\nh2.BinWidth = 0.02;\r\n\r\nxlim([0 1]);\r\nylim([0 1]);\r\n\r\ntitle('Histogram of roundness (probability normalization)'<\/span>)\r\nlegend('coins'<\/span>,'rice'<\/span>)\r\n<\/pre>\"\" 
There you have it. You can add your own object measurements to the output of regionprops<\/tt>. It's especially easy if you tell regionprops<\/tt> to return a table.<\/p>
I'll leave you with this question, dear reader: Are there measurements you would like us to add to regionprops<\/tt>? I am aware of an enhancement request for the Feret diameter. What else would you like to see?<\/p>