PSE 2018 Shake reduction issue

[cz35895897]

When you click on shake reduction is 2018 editor, it goes to work immediately on the selected area of ITS choice. There is no time to move the area or change the size or sensitivity. Moreover it appears that when you do move the area to analyze, it treats the new area as an addition to the shake reduction, as the photo appears to have more "reduction" than before. I would like to have an option to select the area either an automatic or manual, size the area, select the sensitivity and ONLY then perform the shake reduction function. ANy comments on how you handle this issue?

Thanks

[MichelBParis]

Shake reduction is described as 'camera shake', and the video tutorial insists on the common situation of selfies. It's not 'motion blur', move of a subject. Camera shake involves a directional blur for all the parts of the scene; I would expect the filter to find the best location in the image to identify the direction and intensity of the move from the existing sharpness.

You could argue that an image can  include both types of moves; it's a classical situation in sports (cycling, automotive...)

In that case, the 'camera shake' might be enhanced by using layers and masks.

[cz35895897]

I think you illustrate well the issue with the implementation of the tool. With camera shake, objects near the camera undergo a relatively larger displacement due to the angular motion(appears more blurred) while distant object will move apparently less. Unless the camera is moved precisely in a rectilinear motion which is unlikely, and also symptomatic of cycling/automotive/dolly . Hence unless the object of interest to be "unshaken" is the one nearest the camera, PSE will pick the nearest object  with most apparent shake to analyze. However if MY subject of interest is further, or less apparently shaken, it will not be automatically picked. When it works shake reduction works well, with a simple 1. Select area, 2. Commit and process steps it could work even better. (Cost = zero Benefit= large, B/C ratio=huge)

[MichelBParis]

cz35895897  wrote

I think you illustrate well the issue with the implementation of the tool. With camera shake, objects near the camera undergo a relatively larger displacement due to the angular motion(appears more blurred) while distant object will move apparently less.

Tell me where I am wrong.

We both are commenting on an angular move, not a rectilinear 'travelling'.

Imagine two spheres.

One is 0,1 meter diameter at 1 meter distance.

The second, same diameter, is at 10 meters.

The 'shake', that is the rotation, will follow a certain path, say 0,05 meters in your scene.

The same rotation will show the farther sphere with 10 times the path, that is 0,5 meters.

On your print, if the closer sphere moves by 1 degree, so does the farther one.

If the 'shake' creates a .005 meter move for the closer sphere in your print, how long will be the move on your print for the farther one?

[cz35895897]

I think its the trig definition of a sine. Imagine a triangle. The farther away the object is the more it will move. Opposite of what I said before Bottom line various distances from camera will have various blurs for the same angular displacement, so it would be nice not always fixing the worst blur, but the blur of most importance.

However I just ran a small experiment which makes me think possibly my initial take was right. Hold up your finger in front of your face, now turn your head. The finger appears to move more when compared to objects in the distance. Still not certain whether this was valid, as my eye "tracked" the frame, I took some actual photos, with image stabilization turned off. Tried 1 sec, 1/2 sec and 1/4 sec exposure while rotating the camera thru about 30 degrees horizontal. Object at half the distance from the camera (relative to further objects) appears to have a much longer blur streak when compared to the object at twice the distance.Could be that the diagram should be interpreted as far objects need to move far more to cause the same angular error? Try that short experiment (took me less than 5 minutes) and let me know your conclusions.