Lines Detection: Use of the Radon Transform

 The classical approach to detect lines in an image is the Hough Transform. Here I'll present another alternative based on the Radon Transform (seen in my Learning Tomography series [see TOC]).

1- Properties of a sinogram

Create a 64x64 8-bit black image and using the pencil (Tool # 15 after selecting Drawing Tools), draw a white dot.

Fig.1: Dot

Run the sinogram.js script by choosing only 90 projections (an angle step of 2°). Here is the corresponding sinogram.

Interestingly, a dot is transformed into a wave (something like the Hough Transform).

2- Using Radon Transform as a line detector

... and more interestingly lines appear as dots in Radon space.

Then, if you search for the maxima with the command Process > Find Maxima (choose 'List', check 'Preview' and adjust the 'Noise Tolerance' to a high value - 100 in my case after normalization-), you'll get the following list composed of three points whose coordinates are:

N   X    Y   

1   86   81

2   143  45

3   175  157

3. Other crazybiocomputing posts

Further readings are available in ...

• Segmentation Series  [Link]
• Image Processing TOC [Link]