{"id":295,"date":"2009-10-19T07:00:36","date_gmt":"2009-10-19T11:00:36","guid":{"rendered":"https:\/\/blogs.mathworks.com\/steve\/2009\/10\/19\/faster-imreconstruct-in-r2009b\/"},"modified":"2019-10-29T08:01:00","modified_gmt":"2019-10-29T12:01:00","slug":"faster-imreconstruct-in-r2009b","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/steve\/2009\/10\/19\/faster-imreconstruct-in-r2009b\/","title":{"rendered":"Faster morphological reconstruction in R2009b"},"content":{"rendered":"
\r\n

We've been working for a while now to make Image Processing Toolbox functions run faster. The R2009b release notes<\/a> mention several performance improvements. We've gotten some feedback, though, that our release notes are pretty vague about\r\n the improvements. I can't argue with that impression. We tend to be vague because performance optimization is a very complex\r\n topic, and it can be quite difficult to characterize performance changes in a way that is brief, understandable, and accurate\r\n for every user's own hardware and data.\r\n <\/p>\r\n

But I've decided to start posting more detailed information here about the performance improvements. I have more flexibility\r\n (and room!) here than we have with the release notes.\r\n <\/p>\r\n

Today I'll tackle imreconstruct<\/tt><\/a>, which performs morphological reconstruction. Reconstruction is a very useful operation that I've written about here before.\r\n For example, see my post from last year on opening by reconstruction<\/a>. Several other Image Processing Toolbox functions call imreconstruct<\/tt>, including imclearborder<\/tt>, imfill<\/tt>, imhmax<\/tt>, imhmin<\/tt>, imextendedmax<\/tt>, imextendedmin<\/tt>, and imimposemin<\/tt>.\r\n <\/p>\r\n

Let me use the opening by reconstruction example<\/a> as a benchmark case.\r\n <\/p>

url = 'https:\/\/blogs.mathworks.com\/images\/steve\/2008\/book_text.png'<\/span>;\r\ntext = imread(url);\r\nimshow(text, 'InitialMagnification'<\/span>, 25)\r\ntitle('918-by-2018 image displayed at 25% magnification'<\/span>)<\/pre> 
The example task was to find letters containing vertical strokes by eroding with a vertical structuring element and then performing\r\n      reconstruction.\r\n   <\/p>
se = strel(ones(51, 1));\r\nmarker = imerode(text, se);\r\ntext2 = imreconstruct(marker, text);\r\nimshow(text2, 'InitialMagnification'<\/span>, 25)\r\ntitle('Output image displayed at 25% magnification'<\/span>)<\/pre> 
So how long does that call to imreconstruct<\/tt> take in R2009b?  I'll use my function timeit<\/tt>, which you can download from the MATLAB Central File Exchange.\r\n   <\/p>
timeit(@() imreconstruct(marker, text))<\/pre>
\r\nans =\r\n\r\n    0.0241\r\n\r\n<\/pre>
That time is about 45 times faster than the same operation performed in R2009a.  Note that I'm running on two-core computer;\r\n      the improved imreconstruct<\/tt> is multithreaded, so the performance improvement would be greater on a four-core computer, for example.\r\n   <\/p>\r\n   
Now let's time gray-scale reconstruction.  I'll make a 1024-by-1024 test image and compute a marker image by subtraction.\r\n       This kind of operation is often used to suppress small peaks in an image.\r\n   <\/p>
I = repmat(imread('rice.png'<\/span>), 4, 4);\r\nmarker = I - 20;\r\ntimeit(@() imreconstruct(marker, I))<\/pre>
\r\nans =\r\n\r\n    0.0201\r\n\r\n<\/pre>
This time is about 30 times faster than R2009a, again running on my two-core laptop.<\/p>\r\n   
Now for some key caveats you should know.  For now, the performance improvements described here only work for 2-D inputs that\r\n      are uint8, uint16, or single, and only when the specified connectivity is 4 or 8.  We'll be working in the future to extend\r\n      the speed improvements to other inputs.\r\n   <\/p>