Bug fix release 16-Sep-2019

[jsachs]

I have posted a new release (16-Sep-2019).

The changes (taken from the update log) are:

Scratch Removal: added progress indicator, fixed some bugs and improved performance significantly.

Updated to ExifTool 11.65.

[Marpel]

Not sure if this latest release was intended to correct the issue related to Scratch Removal as discussed in the 15-Sep-2019 release thread, but I just downloaded and tried to remove the same curved branch as I first discussed, in a single operation.

Again, I shift-clicked at the beginning and end of the branch, and when trying to add control points along the branch, it did the same thing as before (giving me that X shaped configuration). However, what I did discover, after adding the two end points, I shift-clicked about halfway between those two points (then half way again between the three points etc) and it worked as expected.

And, I must say, considering the branch crossed about 25 other branches in the background (so not a plain background like a clear sky), it did a remarkable job.

Just an FYI for those wishing to scratch-remove a fairly long curved object.

Marv

[jsachs]

When you add a new control point to an existing scratch, it inserts between the nearest existing control point and the closest control point adjacent to it. If you click way off the curve, it may not choose the insertion point you want, so a better way is the click between two existing points and then drag the new point to where you want it. There is no way to insert a new point as the first or last point, so to extend a scratch you need to drag the endpoint to its new location and then insert a new point next to it.

I am updating the help file and Transformations.pdf with this information.

[Marpel]

Sorry Jonathan, I guess I am just not understanding this:

The object I am using as a test for this is a branch about 700 px long and curved (like a bow without the string). If I shift-click point 1 at one end of the branch, then drag the cursor to the opposite end for point 2, it leaves a straight line (a long rectangular bounding box) between points 1 and 2. So far so good.

If I shift-click within the bounding box approx 125-ish px up from point 1, I get point 3 and can then move that point onto the branch (so now a "j" shaped bounding box with a straight line between points 2 and 3).

If I then shift-click point 4 inside the bounding box about 125-ish pixels from point 3 towards point 2 (so now point 4 is closest to point 3 and closer to point 2 than to point 1), the bounding box turns into three straight bounding boxes - point 1 to point 4, point 2 to point 3 and point 3 to point 4. So moving point 4 or 3 results in an exaggerated X.

Sorry for being such a pain on this.

Marv

[jsachs]

Not having much luck reproducing this here - can you post a screen shot?

Not sure if this matters, but the insertion order depends on which control points are closest -- not whether you are near the curve (i.e. in the bounding box). But maybe inserting at the nearest point on the curve would be easier to explain and visualize.

The code is similar to creating a mask with the mask spline tool.

[jsachs]

One further thought about what you may be doing...

If you accidentally click (instead of Shift-click) somewhere away from any of the control points on the scratch, PWP will deselect the scratch. If you subsequently Shift-click it starts a new scratch instead of editing the current one. Not sure what you mean by an X.

[Marpel]

Jonathan,

Sorry didn't respond earlier with a screen-shot. I am getting ready for an unexpected trip so haven't checked the forum for a bit.

Anyway, I just did the same exercise again and grabbed 4 screen-shots of various stages of the process. However, I am just attaching the last one to show you what the result is. If you want them all, let me know.

As you will note, there are 4 points - the far right (1), which is the first shift-click, the far left (2), which is the ending point of the beginning drag and drop, the second from the right (3), which is the first point I shift-clicked in the "bounding box", then dragged to it's current location (resulting in the "j" I mentioned in a prior post), and point 4 which is the third from the right. Immediately upon shift-clicking that fourth point, the new bounding boxes appeared. I dragged the fourth point down a bit to separate the boxes so you could see the "X" I mentioned - except it has a bottom leg between 3 and 4, which I did not initially mention.

Screenshot-(4).jpg (91.46 KiB) Viewed 4134 times

[jsachs]

I see the problem now. The 4th point you insert is being inserted between the wrong two existing control points.

Here is the rule PWP currently applies to decide which two control points to insert between.

First it finds the closest control point.

If the closest control point is an endpoint, then the new point is inserted between the endpoint and its neighbor.

If the closest control point is not an endpoint, then the new point is inserted between the nearest control point and one of its neighbors. It chooses which neighbor to use by finding the nearest one to the new control point. In your case this was not the one you intended.

To avoid having the new control point inserted out of order, the safest way is to insert the new control point midway between the two control points you want to insert between and then drag it to where you want it.

Since the current rule is clearly confusing - at least in some cases - I will see if I can't create a rule that is more intuitive.

[Marpel]

Jonathan,

Thanks very much. Your most recent explanation caused the light bulb to turn on.

I had assumed that inserting a control point inside an existing bounding box would merely recognize the two points on either side of it regardless whether they were endpoints or not. But as per your explanation, in my example, point 4 recognized point 1 because 4 is closer to 1 than it is to 2, even though point 1 was on the other side of point 3.

Your suggestion of adding a new point half way between the two points you wish it to be (then dragging it) seems to work fine.

Marv

[jsachs]

For the next release I reworked the method of determining where to insert new points. It does a much better job of guessing where you intend to insert and also lets you add to either end of the scratch to extend it.