I was going through the project lists of Andrew NG from his Stanford machine learning course and I found a project based on recreation of calligraphic characters. To do so firstly we need to extract strokes from characters and I can only understand it in bits and pieces and I cannot understand at all how to implement it in matlab.
I also read the research paper that the stroke extraction method is based on. But I did not understand that too like there were some terms that I could not understand. And not to mention I can't understand just how to implement all the pixel related operations mentioned. I am a beginner in image processing. Ss please any help is appreciated.
Quoting some lines from the paper:
We use the direction contribution of the segments to estimate the degree for each pixel on the thick-line image. Given a binary image, for each black pixel $p(r,c)$, let $D_k(r,c)$ denote the orientation distance between the pixel and the boundary point along the $k$th quantized orientation, where $k = 1, 2, \dots, M$, and $M$ is an integer that denotes the quantization number from $0^\circ$ to $360^\circ$ . The orientation distance is called point-to-boundary orientation distance (PBOD). The distribution of the PBODs contains information about the degree of each pixel. We can easily tell that the PBODs along the quantized orientations of the branches are much larger than that along other quantized orientation. So, to estimate the degree of each pixel, we need only calculate the number of the crests of the distribution of all the PBODs of the pixel. Figure 2 shows some cases of the distribution. The resolution of quantization is $3^\circ$.
What is quantisation and orientation distance, and how to implement these in matlab?