I’m reading trackmate documentation recently. It’s a great work. I have some questions when reading it. In spot location part, it mentioned that trackmate uses a quadratic function, but not to much detail. I went to trackmate github for the code, but I didn’t find quadratic fitting. Can someone show me where the quadratic fitting part of work(github address)? I also read the reference mentioned(Lowe 2004), about spot location. To generate the scale space, what is the kernels’ scales trackmate used? Are they σ ( σ = r/ √ n) , 2 σ, 3 σ…? And how many kernels for each octave?
I simply reused an ImgLib2 function made for scale space detection. But there is actually no scale space in the LoG detector (just one scale, the one you set by inputting the blob diameter).
You can find the code here:
Thank you for your reply. You mean trackmate directly use LoG detection result to subpixellocalize the spot, which is L(x, y, σ) = G(x, y, σ) ∗ I(x, y), here σ is r/ √ n? Any down sampling applied to the image?
20 char limit mandatory sentence.
Got it. Thank you for your help.