Exported image quality
When I export my D810 RAW files to jpg I typically have 900 px set for the longest side and the quality is set at 100.
The D810 files are 7360 x 4912 so this is a x8 reduction.
I have found that the perceived resolution of these jpgs when viewed in an image browser is very poor.
In photoshop there are resampling operations that are specific for when reductions rather than enlargements are made. In addition Photoshop has sharpening filters that work better when doing reductions. What are the equivalent processes when working in CaptureOne?
Ian
The D810 files are 7360 x 4912 so this is a x8 reduction.
I have found that the perceived resolution of these jpgs when viewed in an image browser is very poor.
In photoshop there are resampling operations that are specific for when reductions rather than enlargements are made. In addition Photoshop has sharpening filters that work better when doing reductions. What are the equivalent processes when working in CaptureOne?
Ian
0
-
900px is an odd size to choose.
Any particular reason for it?
I'm not sure how the browsers and viewers would deal with it.
If you take a Photo Library system such as, for example, Flickr one finds these standard sizes, presumably chosed for specific reasions.
Square 75 (75 x 75)
Square 150 (150 x 150)
Thumbnail (100 x 66)
Small 240 (240 x 158)
Small 320 (320 x 211)
Medium 500 (500 x 330)
Medium 640 (640 x 422)
Medium 800 (800 x 527)
Large 1024 (1024 x 675)
Large 1600 (1600 x 1055)
Large 2048 (2048 x 1351)
Based on an original of 5921 x 3905
Some web sites, when displaying posted images, expect 650 or 750 to fit their format sizes.
Going outside that may mean they manipulate the image internally to make it fit and that rarely does it any favours.
The general consensus seems to be that if reducing the file size when re-sizing it is likely to beneficial increasing the sharpening effects. Fine sharpened outlines will tend to be somewhat lost when 7/8ths of their pixels will be discarded although no doubt the selection of re-sizing algorithm will have an influence on that.
There is then the issue of how the viewing screen deals with the image size compared to it native dimensions as currently set. And then there is the subject matter which may or may not be especially prone to changes based on display formats.
Not a simple question to answer in all cases as far as I have found.
Grant0 -
Grant,
Nikonians forum until recently asked for 900px max for longest side upload. That has now been increased to 1800px
You will note I only specified the longest side and both Photoshop and CaptureOne pro allow this.
My question was how to avoid significant quality reduction of resolution when downsizing. All the aspects you listed do not answer that question.
Ian0 -
Don't downsize in Capture One - I've pointed out before on here that it does a poor job. 0 -
[quote="Beemer" wrote:
Grant,
Nikonians forum until recently asked for 900px max for longest side upload. That has now been increased to 1800px
You will note I only specified the longest side and both Photoshop and CaptureOne pro allow this.
My question was how to avoid significant quality reduction of resolution when downsizing. All the aspects you listed do not answer that question.
Ian
Ok, that's up to the web site of course. I think it may suit the sensors but not necessarily the resolutions of all monitors. The bigger the sensor the more information would have to be thrown away - presumably that is why they have changed the maximum size allowed. Not much point in having full size images in glorious detail crushed to suit the internet.
Grant0 -
Grant,
Your comments are quite puzzling. You said (in relation to my outputting 900 px) that it may not suite the resolution of some monitors. Would you care to elaborate?
Image pixels are sizeless and the resulting display size depends on the ppi of the monitor. It is only if the image browser interpolates to fit the image to all of the screen and fails to match the monitor native pixel count that FULL SCREEN image quality suffers.
However in my case a 900 px longest side image is nowadays much smaller than the typical screen ( mine is 1920x1200) and the fact that this is not a "preferred" size should not result in poor image quality.
Ian0 -
Ian, you are correct.
There is an issue with how CO downsamples/sharpens reduced size jpegs.
viewtopic.php?f=57&t=18817
viewtopic.php?f=57&t=18271
viewtopic.php?f=57&t=177350 -
interpolation of diagonal lines can be particularly challenging when compressing detail. In general, to my understanding,
scales that represent pixel arrays in ways that are mathematically ideal for computing purposes tend to offer a better opportunity for good looking results when infilling.
Once you take into account requirements for, for example, posting into existing forum screen dimensions and then, perhaps, scaling the entire screen to your personal viewing preferences (perhaps to read the text comfortably) things become a little less controlled.
I have a a 1920 x 1080 22 inch screen for the desktop and a 15.4 in screen on the notebook running the same pixel resolution but obviously in a much smaller area. Depending to some extent on viewing distance and and light output settings the detail in a given image may look good or not so good from one screen to another - if I was being very critical.
In my case both screens are more or less comparable in technology terms but that is not always the case.
Looking at pixel density the larger screen calculates out to about 101 pixels per inch using the full screen area although of course one might leave a screen border and so match closer to the typical 96ppi that is commonly referred to for reference scaling.
The notebook screen, at full area utilisation, suggests about 142 ppi are available.
Whether the image will look identical, or even comparable, on both screens may depend on a number of factors not the least of which would be out own eyesight and viewing distance.
The nearest quick test I can think of would be to zoom the 142ppi display to about 150% to approximate a 101/96ppi display. At 150% any flaws in straight boundary edges, vertical, horizontal and especially diagonal, are becoming evident in a way that is not obvious at the smaller size, Conversely if one goes the other way these details and interpretive details form the processing as lost in the compression. Quite good for hiding problems in large areas, not so good if seeking fine detail retention.
Basically my Notebook screen has smaller pixels and it is unlikely that my eyes at anything like normal use distance, absent a magnifying glass, can discern the pixel to pixel changes in detail.
With a larger screen of the same pixel dimensions the pixels are larger and even if viewing from somewhat further away the potential for distinguishing individual pixels is greater.
In addition the screen may not be set to use all of the available pixels (i.e. not exactly edge to edge or within the measured pixel area) so some interactive interpolation could be in play for scaling purposes no matter what you have achieved with the output file.
It all gets a bit complicated but I think the general feeling is that binning 1 pixel in 4 has been acceptable down the years and a decent compromise on achievable quality vs size whereas binning 1 in 8 (plus a slightly significant bit of another somewhere) is not so helpful in the quality stakes.
Maybe we realoly need some worked examples here - perhaps even an assessment of various browsers too.
Grant0 -
This has nothing to do with 'ideal numbers/ratio' for down interpolation or secondary downsizing in a browser/viewer or screen pixel pitch.
The bottom line is that CO does a crappy job of downsizing with poor anti aliasing and/or excessive sharpening applied.0 -
[quote="Alex1111" wrote:
This has nothing to do with 'ideal numbers/ratio' for down interpolation or secondary downsizing in a browser/viewer or screen pixel pitch.
The bottom line is that CO does a crappy job of downsizing with poor anti aliasing and/or excessive sharpening applied.
Not something I have noticed unduly with my images - only other peoples'.
Maybe I don't have enough pixels to make it a visible problem when compressing.
Grant0
Post ist für Kommentare geschlossen.
Kommentare
9 Kommentare