Step Wedge without boundaries

[tomczak]

Trying to create 'pure' step wedge, I was able to turn off the labels, but I couldn't take out the dashed boundaries between the patches and around the image. Patch spacing stops at 2.

The reason I thought it useful was that I was trying to explain to myself how exactly the waveform histogram works by trying it on different test images.

[tomczak]

Then, when I thought I had it all figured out, I used Q13 image, cropped the 50% gray boundary and made horizontal and vertical versions of it, expecting 20 (the number of patches) clean steps for horizontal image and 20 clean horizontal lines for the vertical image, the waveform histogram surprised me again...

[tomczak]

Another interesting example is the image of Colour Wheel in HSV-V, which has all the V values set to 100%, except for the black boundary. The Waveform is mostly black, even at full expansion, which I imagine is the result of colouring brightness for this mode, but how the colours cancel each other out?

[tomczak]

Try these two patterns I've made - I almost understand why the waveform line is yellow (it's minus blue), but what are the artifacts at the colours interface?

[jsachs]

Note that the entire waveform is scaled so its largest value corresponds to 255, so the solid borders around the Q-13 generate large spikes that scale everything else down. If each vertical histogram was scaled independently, they would not be consistent with each other. You can overcome this to some extent using the expansion slider.

[tomczak]

While this is just an artificial image that I'm using to learn, I can't figure out why those multiple lines exist. The colored waveforms are from green/red images above (vertical and horizontal), the B&W one is for extracted HSV-V channel of either of those images. I understand the B&W one, I think.

[jsachs]

When you crank expansion up to 5, single pixels will show up, even though they are not significant in how the image looks. The code that computes the waveform first resizes the image the get a reasonable number of pixels to work with. In the examples you are using, this can create some false colors at the transitions between blocks of solid color and these are showing up as extra lines when you expand the histogram. On a normal photographic image this would not be an issue.

This video may be of some use:

https://fstoppers.com/education/out-his ... rms-346898