Greetings. From time to time I study some aspect of processing. This image is the result of something I'm currently working on involving preservation of detail in large crops...(a bit of a long story). Here's the processing on this image of an Oak Titmouse eyeing my backyard feeder.
Cropped 88% of the pixels for a 1.7 MP image.
Two NR passes on whole crop then selectively on bg.
Uprezed by 300% w & h.
Detailed with Topaz Adjust first, then Topaz Detail
Nik Silver Efex Pro for contrast and structure
Topaz Simplify BuzSim, small simplify size, monoline fine edges
Reduce to BPN sized tiff shown here (I was quite surprised by the appearance of the fine feather detail after the size reduction)
Export to jpeg
The feather detail is great. I would not have expected that after two NR passes, buz sim, etc. You really used each filter thoughtfully and very effectively creating this image. It is good you posted your routine as I want to think about it in detail. The orange-red bkgd sets the titmouse off very well. I wonder if you want to bring that color into the bkgd that is under the perch?
Hi, Michael, I love titmice! They are regulars at my feeders but they're pretty speedy. I'm surprised too that you were left with so much detail. Thanks for your workflow - I always learn a lot from your images. Nancy makes a good suggestion about the area underneath the perch.
"It is only with the heart that one can see rightly" - The Little Prince
Thanks much for the comments. They are much appreciated.
So, I've been working on understanding why one can tell in a 1200x800 BPN image whether it was from a 2400x1600 crop or a 4800x3200 full frame image. Because one can tell. I've seen it in my own images and those posted on Eager to Learn and Avian as well. But a 1200x800 image can only hold so much information, how is it that it can hold information to distinguish if it comes from a cropped original or a full frame original? My professional background in synthetic digital image processing says it can't hold such information... so what gives?
After studying this I believe there is a, uh, well artifact to the resizing step that I'll call "shrink to detail". This image is an example of how that works. The image I've attached here shows a 100% pixel view of the uprezed version of the image. If you have Topaz Simplify you'll recognize the characteristic output of buzsim, here sharpened a bit ending in a somewhat posterized image.
The only steps between the OP and this image (3546x4433) is reduction to a 720x900 image using bicubic sharper and output to jpeg (from 16bit tiff). The fine feather detail in the OP jpeg originates with the posterized lines in image in this panel. The thick posterized lines when shrunk (with some help from bicubic sharper) end up as sharp detail lines. Cool, huh?
I've been experimenting a bit with this as a straight detailing method for cropped images (not the buzsim part just regular detailing but on uprezed versions before downsizing). Seems to work pretty well.
About Noise Reduction: noise reduction is important before uprezing,,, uprezed noise is too big for noise reduction tools to do their thing.
I'm working on understanding this but it may take some time. Just got interested in astrophotography and am getting ready for a 2-week trip, so I may be a slow learner here.
Your words, "But a 1200x800 image can only hold so much information, how is it that it can hold information to distinguish if it comes from a cropped original or a full frame original?" made me think this was some limitation of the resizing algorithm. I'd bet it's way "simplified" -- OK, pun intended.
Maybe this is worth starting a new thread in Digital Photography Workflow?
Detail Cartooning From Large Crops, Oak Titmouse
