4
votes

I'm new to MATLAB, and this is the problem I'm facing.

I have many spiral RGBimage like a galaxy, cyclone, bisected conch, etc., and I'm trying to fit a log-spiral on to it. However, the accuracy of fitting the spiral I'm getting is very low, and I have to increase the number of points in the plot so as to increase the accuracy of fitting.

EDIT: The input images are similar to the images in the below link (I have manually drawn a log spiral curve), but made a curve to fit and it's not fitting properly.

I.e. a curve is to be fitted as shown in the below image, automatically. enter image description here

WHAT I DID: Done some preprocessing and morphological analysis to get only the body or the main spiral part of the image.Detected the center and 2-3 points on the spiral using Image processing.

But I need more points to fit the curve perfectly, so i need to convert the initial image to a image with more dots/points on the spiral as shown in the below link but in 2D. i.e. get points instead of white strip of the tornado which is in the form of spiral, so that the curve fits perfectly. I need to convert the spiral image of tornado to a scatter plot similar to the below.

enter image description here

Thank you.

1
Can you provide example of input image. - Marcin
Your question is not really clear, please provide more details. E.g.: How do the images look like, how do you currently fit a spiral to an image, an example where the fit works, another where it does not (as well). In the last sentence it seems you introduce a completely different approach, difficult in itself – why do you think that's a good idea? Etc. - A. Donda
RGB image cannot be converted to a plot, as far as I know. The figure which you have shown can be generated with the command scatter3 - Autonomous
Nice pictures but really difficult to understand. What is a conversion to "an image with more dots"? What does "get points instead of white strip of the tornado" (which tornado) mean? Especially tornado confuses me, I don't see one. - Trilarion
Some code would be nice to see - Hennadii Madan

1 Answers

0
votes

I believe weighting is the answer. You said you did some image processing to get you spiral arms. Now use the whole image as an input to your regression, but assign weights equal to intensity of each pixel in the processed image.