Hi
@JeffK
I tried several approaches to see if I could isolate the jigsaw puzzle lines. It turned out it was not very complex (and I realized I could have taken out a step as well). Here are the steps
I found the best online image of the puzzle with the image Google search feature. (following are all screen shots)
I used the neural filter to remove JPEG artifacts and then inverted the image.
I applied a Threshold Adjustment Layer and moved the slider to the high end, where I had a good balance of the jigsaw pattern not disappearing yet eliminating most of the underlying image.
From here, I painted black on the few areas that were not puzzle pieces and also added and painted white in the few areas where the puzzle lines were broken. Accuracy was not important.
I selected the white to create the selection and increased the selection size by two pixels so that the selection overlapped into the image and left none of the puzzle lines. Then I just ran content-aware fill to achieve this first image with fuzzy out boundary:
To put back the outer edge, I just took the original image, created a boundary selection around the outside edges of the image, and used the inverse of that selection to create a mask so that the edge areas were seen from the original image below:
I did run the dust and scratches filter to eliminate such spots here and there.
Hope that helps
John Wheeler
