Help - Smoothing surf plot, reduce surf detail and speed up performance
조회 수: 106 (최근 30일)
I've got a large 3000x2000 matrix (Z) that represents the depth values for a topogrpahical surface plot. Because of the blurring between colour depth I get a very pixelated, sharp ridge line (see image 1 below). I've applied a medfilt2 and then interp2 to try and smooth out the ridges on the plot. This seems to work, but it's a lot of data points and it's running very slow. Is there a way to reduce the points and smooth the plot?
Z_med = medfilt2(Z, [5 5]);
Z_int = interp2(Z_med);
s = surf(Z_int);
% This gets the best results, but is slow and hard to work with
I've tried plotting less points (every 10th point), but this just resulting in really jagged edges (see image 2), and I'm not sure if the way I did this is right (it doesn't look right)? I've also tried plotting a courser grid, but the way I am doing it is wrong?
% Method 1 - Jagged ridges - Failure
step_plot = filt_plot(1:10:end,1:10:end);
Z_int = interp2(step_plot);
s = surf(Z_int);
% Method 2 - Trying to interp2 for a coarser grid - Doing something wrong here - Fails
X = size(Z_med,1);
Y = size(Z_med,2);
[Xq,Yq] = meshgrid(1:2:X,1:2:Y);
Vq = interp2(X,Y,Z_med,Xq,Yq,'cubic');
I was hoping to get a smooth plot between depth layers, and plot less points, but not getting the jaggered spikes like below...
Any help would be greatly appreciated!
추가 답변 (1개)
Deniz Diktas 2020년 11월 18일
I am answering your question late in the same year. The rendering of topographical data (in other words, terrain data) is not trivial. Downsampling is the way to go but there are issues with how you do it: you should not leave out arbitrarily selected sample points when downsampling. By arbitrary I even mean regular downsampling. Such sampling may mislead you visually. The correct way to go about this is to render the data depending on the view-point (both camera location and view direction), which is a non-trivial issue. In 2014 I worked on a project for real-time terrain rendering and I can easily say that this is an intermediate or even an advanced topic in real time computer graphics. The main idea is to divide the data into smaller tiles and produce multi-resolution versions of these tiles (downsampling according to certain criteria comes at this stage) and then sending these tiles at appropriate resolutions depending on view-parameters in real-time to the GPU. I wonder if such a rendering is supported in Matlab's mapping toolkit.
Let me know if you want to learn more..