Improve transform bounding box handle accuracy, improve canvas navigation

[HumbleDeer]

The handlepoints ("corner balls") used to adjust the transformation window ("bounding box") have a couple issues:

• The handles and bounding box change size alongside the zooming of the canvas. This mostly negates the reason to zoom in, which one would do to increase their accuracy in placing the bounding box.
• The handles and bounding box lines can obscure the image you're trying to place the bounding box on.
• Clicking the handles to drag the window causes the window's corner to assume the location of the cursor. This causes the bounding box to "jump" away from the previous location when trying to fine tune its location.
• Using middle click to move the canvas only works when the middle click is done within the image's bounds. Middle clicking on the grey background does not move the canvas.

To improve upon this, either indirectly or directly:

• Maintain size of the handles and the bounding box when zooming in. This results in a simultaneous increase of the precision of the adjustment as one zooms in. Seems intuitive, as that's why you'd want zoom in.
• Add a "loupe" overlay that shows a zoomed-in view of the area where the handle is being adjusted in. The overlay should not have any of the handles showing inside, so that the user can accurately gauge the location.
• Have the handle dragging/clicking logic tolerate the cursor being anywhere within the handle's bounds to drag it around, without the bounding box jumping to that location.
• Allow the user to move the canvas when middle clicking anywhere within the canvas (including grey background).
• Add the option to restrict moving the bounding box handles beyond the image bounds (handy for adjusting the box to something that runs off the side of the image (plus toggle)

Tangentially related suggestions:

• Change the cursor to the "grabbing" state when the canvas is being moved with the middle click button.
• Add a "status bar" type zone at the bottom of the window where controls for the canvas can be located:
  • Button for reset zoom
  • Button for fitting to the window
  • Button for zooming to fit the transform bounding box to the canvas, with some margin
  • Toggle for allowing / restricting bounding box handles placement outside the image bounds
  • Zoom selection on a slider (to indicate as well as navigate)
  • Status of the processing: Could range from just indicating processing is happening to also showing feedback about the stage or the percentage progress.

[HumbleDeer]

I've edited the message with some additional points

[TheJoeFin]

I have been planning on revamping the canvas with this package: https://github.com/mircea21S/RichCanvas

In addition to that work I want to implement many of the changes you are suggesting here. Thanks for being so thorough.

[HumbleDeer]

You're welcome!

I ended up using a plugin for Paint.NET which provided the features needed but still with a suboptimal experience, so that got me wondering whether PDN would allow hooking in an external program in its entirety to allow an external window to provide features such as the boundary box dragging.

For references on good UI/UX of adjustments like the corners and the transform, the PhotoScan app from Google always did a fine job for me. Google Photos used to have this feature too, but recently removed it.