flagpole wrote:It's interesting the different ways you've all handled the bottom of the spine of the book.
True. :j-)
I left the spine on the book in my first post when using the smaller posted image and also did a version with the larger image like that but didn't post it. I actually believe the bottom of the spine does belong there because, as I said, I believe it is the spine of the black cover and not shadow from a close look at it. But I figured I'd remove it with the second post I did to most closely resemble hartpaul's result but using the selection tool in two ways - both Magic Wand and Freehand - and to indicate that no brushes or erasers for cleanup are needed to do the job even around curved parts virtually the color of the cover.
flagpole wrote:It's interesting the different ways you've all handled the bottom of the spine of the book.
True. :j-)
I left the spine on the book in my first post when using the smaller posted image and also did a version with the larger image like that but didn't post it. I actually believe the bottom of the spine does belong there because, as I said, I believe it is the spine of the black cover and not shadow from a close look at it. But I figured I'd remove it with the second post I did to most closely resemble hartpaul's result but using the selection tool in two ways - both Magic Wand and Freehand - and to indicate that no brushes or erasers for cleanup are needed to do the job even around curved parts virtually the color of the cover.
Yes I didn't like to say you were all wrong
I believe it's like a double spine and a dark shadow. Should look something like this
The original issue I posted about was ensuring the edge of the image would anti-alias properly with any new background. If you use any method to cut out the image with any degree of softness you can still get pixels at the edge which are anti-aliased with part of the original background which gives them an HSL which can be inappropriate for any new background, or you go too far into the original object and lose edge details.
The soft edge to the eraser and background eraser is one solution but I find it can be too soft and leaves the image looking fuzzy.
The example I uploaded is a fairly straightforward one (apart from the shadow) but I will also have to be able to deal with objects which have a very mixed background.
My current solution (based on the suggestions here and my own tests) is:
1: Enlarge the image by 300% [EDIT- If it is not already a high resolution image]
2: Create a duplicate on a new layer
3: Use the Edge Preserving Smooth option on the duplicate to flatten colour areas so you can use the magic wand to select and delete them more easily. [EDIT - my images are from film so have a lot of grain]
3: Use the brightness and contrast to isolate the required element as much as possible (don't worry about losing detail as the original detail can be recovered from the original layer at the end).
4: You may need to fill or paint the edge of areas which merge with the background too easily or you may need to use the magic wand, eraser brush or background eraser to remove any elements of the background which won't go away using the above techniques but the work is greatly reduced. Here you can also use the other suggestions recommended such as the object extractor to get at the core object. The key is to use a judicious combination of edge-preserving smooth and the brightness contrast.
5: Once you are confident the object can be selected easily, resize the image to the size you want to use. If you don't do this at this point the anti-aliased edge we are about to create will be affected by the resizing algorithm.
6: Use the magic wand or selection tool to create a selection of the object. No need to feather unless that's the technique you've decided on for the soft edge (see below)
7: Switch to the original layer with the selection still intact - you can now cut the original object out neatly by using Copy, and then Paste as a New Image. If you find there are still artifacts at the edge you can contract the selection by a pixel or two to remove them.
8: To ensure the edge of the new object is not jagged simply use the Shape-based Anti-alias feature on the Selection-Modify menu on the new object. If that doesn't do what you need use the Select Selection Borders feature using a 2 pixel width and then apply Soften to that selection region. As suggested here, you can also create a 2 pixel feather to the selection before copying it but I didn't find that quite as successful as the Shape-based anti-alias.
Hope all of the foregoing (including everyone else's suggestions) helps anyone in the same position. All images are different so you may need to use a combination of these techniques for best results on the image in question.
Last edited by PluginCreativity on Fri Jun 03, 2016 10:20 am, edited 1 time in total.
PluginCreativity I admire your work ethic. But It's way more than I do to get the required results. In the below example I have cut out the motorbike. Wheels, spokes and all. All I do is use the Freehand Selection tool set to Point to Point. Feathering and Smoothing at 0. Anti-alias checked. Zoom in to 300% and cut away. (I do it in small sections to avoid an OOOPs moment). (My files are 300ppi and 6000*4000).
These files are then printed as A3+ and sold in top end museums etc.
All my edges show no sign, or very minimal, jaggies.
Same results for the many vehicle files I also do.
brucet wrote:(My files are 300ppi and 6000*4000).
These files are then printed as A3+ and sold in top end museums etc.
This is OT, but just curious about whether you do the printing yourself or have them done at a commercial printer. And in that regard, I guess I should ask if they are sold as photo prints, silkscreen, or some other output medium.
Thanks guys.
I do my own printing. I'm a control freak!!!!!! (Re prints MUST be the same as earlier prints).
Hence I mix the DPI's and my PPI's. At the end of the day as long as everything is 300 something I ignore the DPI v PPI debate. And get into trouble for it!!
I use PL and PSP to do my photos. Print them using QImage on an old Epson 1410. Use Moab paper. Mount all the prints myself and supply them to museums ready for sale in plastic show bags. (Have been known to have 100+ prints in my suitcase and flown across the Pacific to do a delivery). Prints, licenses etc now total over 4.1 million. But I'm bragging. Again. Sorry.
Any chance you could post a before and after close up of a portion. Your suggestion is much quicker and results in a clean image but, for me, it loses detail and sometimes still retains artifacts which need further work.
brucet wrote:... Prints, licenses etc now total over 4.1 million. But I'm bragging. Again. Sorry....
Impressive numbers - you have every right to brag IMO
What we really need of course is vector node editing of the resulting selection and then we can tweak the result whatever tool we use to get started - or have I missed that feature too?
My originals are on another computer so I'll see if I can post you a before and after.
Keep in mind that no one method will work for everything. Sometimes I do use a low Smoothing and Feathering setting. Horizontal and vertical lines also require less work. But even with all the knowledge and skill there are still clean up jobs to do. It's very subject dependent.
Another 'trick' is to use the Smudge Tool. Set it to Size 2 or 3. Hardness 50. Play with the Opacity. (I use around 60). Zoom way in and slowly tidy up any offending edges. (Use the shift key to get straight lines).
At the end of the day as long as everything is 300 something I ignore the DPI v PPI debate. And get into trouble for it!!