Changing image rendering?

Hey All,

attempting to visualize a raster of lines and running into issues. When the lines are plotted as lines, the image appears correctly, as follows:

When attempting to visualize the same lines (ignoring the large vertical and horizontal lines), I get the following

zooming in, all the same lines are present, just zoomed out they have varying degrees of visibility. Is there a way to modify MATLAB handling of this image to produce an image closer to what appears above?

[Oliver Woodford]

I'm just wondering why, if the results using plot are correct, or what you want, why you're trying to rasterize the lines. What is wrong with the approach using plot? If it's rendering time, please share how you call plot.

Answer 1

[Oliver Woodford]

Basically you are rendering a binary image, and you're getting gray levels instead of just black and white?

What are you using to render the image? imshow() can rescale the image if it is too large. imagesc() doesn't do this.

For example:

>> A = rand(50, 4000) < 0.99;
>> imagesc(A);
>> colormap(gray(256));

when zoomed in gives:

[aboharbf]

It is already binary. they are arrays of roughly 5 - 20 rows with ~4000 columns, every entry is 0 or 1.

[Walter Roberson]

spy(data) ?

Anyhow, what controls are available are described at https://blogs.mathworks.com/graphics/2015/07/13/graphicssmoothing-and-alignvertexcenters/

[Oliver Woodford]

See my updated answer.

[aboharbf]

So the issue isn't when it is zoomed out, my hope is to recapitculate the behavior of drawing tick marks w/ plot() using an image drawing function instead, since im drawing many hundreds if not thousands of ticks.

I was using imagesc(), and I realize the image isn't 'big enough', but I wanted to know how to drive the behavior to get closer to the fully drawn line way. It does seem like your sc functions improved things. I will investigate the smoothing and vertexcenter function next. Thank you both!

[Walter Roberson]

You get gray from binary when MATLAB applies anti-aliasing, especially when you are trying to render more data pixels than fit into the physical pixels.

[Oliver Woodford]

It's not possible to do using any of the builtin image rendering functions. They will either blur or do nearest neighbor downsampling, in which case you get gray or lose lines. The plot function renders a line, however zoomed out you are. This is akin to resampling using the min() operator. What you can do is resample in the x direction by taking the min over blocks of columns (or max if your sparse lines are represented by 1s, not 0s).

[aboharbf]

Walter - I attempted to find the 'AlignVertexCenters' property of my figures but it doesn't exist. Not sure why. I'm using 2019b.

Oliver - Would this need to be setup via a custom callback or are there any methods available to do something like this?

[Walter Roberson]

AlignVertexCenters is a line property, not a figure property and not an image property.

[aboharbf]

Then in this case, it seems like it would defeat the purpose, since I'm trying to get rid of the greys present in an image as a way of avoiding the much more computationally taxing task of drawing many lines. It is a handy thing to know about though, so thank you for sharing.

[Oliver Woodford]

I was suggesting shrinking your image to the resolution you intend to visualize it on screen, before passing it to the rendering function, rather than have some kind of callback which actively resizes the image to fit the zoom. Do you need to have the ability to interactively zoom in?

[Walter Roberson]

The features I pointed to are the only graphics smoothing and anti-aliasing available. In particular the GraphicsSmoothing figure property would be the one applicable.

[Walter Roberson]

Maybe conv2() the array with ones() the size of the number of data pixels that will be rendered down to one physical pixel, and logical() the result. That would be equivalent to an OR operation over the entire block.

[aboharbf]

@Oliver - I do need the ability to interactively zoom in and out.

@ Walter - I tried the Graphics smoothing but it didn't appear to have any influence unfortunately. Likely due to the plot being purely vertical lines. The function you describe may be worth a shot though, thanks for the tip.