Too much going on? C&C requested

[tameigh]

Part of the sharpness issue is that the image did not crop to an even divisor of the total that is near 720. I divided it down to 972, which is an even 1/4 of 3888 which is the default size of the raw shot, so the image got crunched by the host site down to the size of this site, which muddled it up a little bit.

I was thinking about cloning out the branches over the snow, I did a little clean up there. And I was wondering about the ice at the bottom. It deserves a macro lens in and of itself, but that does not show up in this pic.

Finally, the blue shadows, I like the blue with the gold of the moss, but that is now how it appeared to me in real life, and yet still I thought it was worthy of a picture. I guess I should start with real life, and work from there.

[Dave Humphries]

Part of the sharpness issue is that the image did not crop to an even divisor of the total that is near 720. I divided it down to 972, which is an even 1/4 of 3888 which is the default size of the raw shot, so the image got crunched by the host site down to the size of this site, which muddled it up a little bit.

Hi Tim,

If you're using Photoshop/Elements/Lightroom, or probably any decent image editor these days, I'm not sure the concept of downsizing in divisible number of pixels 'holds any water' these days.

It is my belief that these programs all up sample into RAM by a factor of 10* (a strong clue being the ability to sharpen (and many other things) in "0.1 px" increments - think about it !

* then downsize again when saving - this is all invisible to the user, of course

For ultimate sharpness of an attached or album image here at CiC;
You must down size to 700px (or less) ideally from an image at least 3 times that pixel count on the longest edge (e.g. > 2000 pixels)
You must re-sharpen after the down size at about 0.3px and 70-100%, using as low a threshold as image noise allows; 0, 1 or 2
You must save as jpg quality 8 or 9, more is a waste of space and may take the image over the file size allowed

If you saved this at 720 and the forum downsized to 700, that will have harmed it (assuming it was sharp enough to start with)

HTH

[tameigh]

Well, thanks for the advice on formatting for display here, but I think that this is one of those images that will never look good on the web. It looks a lot better in photoshop. I still think the divide by four thing works, but this is not a good image to prove it by.

Edit: Now it looks even softer :·(

[MiniChris]

I am going to jump this thread for a moment and do as Dave suggests in setting up an image for display here...something either didn't translate correctly, or I'm nuts...and either is a good possibility.

This is as entered in the competition today


This by Dave's suggestion to downsize first, then sharpen

[MiniChris]

The second one to me looks slightly pixelated and I seem to have lost some of my better midtone values...What say the rest of you?

[tameigh]

I agree with you Chris.

[MiniChris]

Here are two more. The first is downsized then sharpened, and the second is sharpened, then downsized. In photoshop, with the ability to blow the image up, there is a vast difference between the two images, with the first (Dave's way) going pixeloutofsight almost at the first 2x and the other holding much longer 4 to 5x.

However, in fairness, there seems to be some advantage to downsizing and then sharpening as per mechanical devices go, but not as per skin values...hmmmmmmmm

A Note: I use the LAB Sharpening technique and these were done at 125%, 1 radius and 3 levels

Sharpened after downsize


Sharpened prior to downsizing...

[jeeperman]

You might have lost some midtone {very slight} But Chris look at the facial features and the fishing line on the second image....definetly sharper in the second image. First set.

In the second set the first image shows more detail, the dial on the reel and the scuff mark on the frame. It is however a bit darker. Although I am not looking at them full size.

[MiniChris]

I think I have found a way to have my cake and eat it too...downsize and sharpen using LAB method, the save to web and then resharpen using Luminosty method...I am still playing but....

[Geoff F]

Well, thanks for the advice on formatting for display here, but I think that this is one of those images that will never look good on the web. It looks a lot better in photoshop. I still think the divide by four thing works, but this is not a good image to prove it by

I often get similar problems, Tim. Sometimes a photo will print fine with good sharpness and contrast but I am not satisfied with how it appears on the internet, although I have (to the best of my knowledge) done everything correctly.

When doing a substantial downward resize, I sometimes do it in two stages (but I'm not really sure if that makes any discernible difference). Then apply a little Unsharp Mask to suit. I don't think there are any hard and fast rules for the amount of sharpening except, 'Don't overdo it'.

I have tried using the often recommended method of using Bicubic Sharpener for downsizing but I found that the auto sharpening sometimes didn't give the best result. Which is why I returned to a little USM after resize; and just use the standard Bicubic method.

My preferred method is to set the required sizes from the Image Menu then, if there is a required min/max file size, check the file size and set a compression value to suit using Resize for Web; or whatever system your software uses to show file sizes.

Incidentally, CinC has a 150 kb max file size (I think that is correct) but I appear to get better results if I can get it below 100 kb, although I can't offer any logical reasoning as to why.

The resizing by exact percentages was certainly the recommended method some time ago but, as others have mentioned, the current wisdom (which is produced by cleverer people than me) reckon that method isn't necessary now. It certainly makes resizing a lot simpler.

ps. At one time, a method similar to Chris's example recommended sharpening on a layer which was set to Luminosity Blend. Most times, I never noticed any improvement upon the simpler method, although sharpening of selected areas using layers can be very productive. Perhaps I may now have another think about this method.

[tameigh]

Thanks Geoff,

Where the luminosity sharpening really shines is when the picture contains a lot of tree branches, because they halo so quickly under normal USM, but other times, the difference is extremely slight, but I have never seen it produce an effect of zero.
- Tim

Works nice on fish scales too.

[MiniChris]

I've found it works best with shiny things and liquids...but it also has a way of smoothing skin tones out without losing the color value. A luminosity layer only affects the luminosity of the photo and not the color data. It is one of those techniques you have to experiment with and I spent several hours testing different photos to see which ones showed a defined better image...most times, I just use LAB and channels.

[Dave Humphries]

Random quick replies:

The aim was to make an image that is sharp when viewed at 100% and at 700px on longest side in a browser with no zoom/magnification - that's all - so blowing it back up again in Photoshop to 2 x or 4-5 x is obviously asking for trouble.

Also I don't use LAB, but 125%, 1 pixel is going to be noticeable if applied to a downsized image (= overdone, in my view)

I don't suggest recommend "bicubic sharper" either, for the reasons Geoff says.

Bear in mind that if you sharpen at a given radius, it'll also be divided by the downsize ratio, so; a 1 pixel radius on a 2100 pixel image, then downsized by by a factor of 3 to get 700px equates to about 0.3px radius, except it won't be as sharp as if done after the downsize at 0.3 radius because of noise on edges and other stuff I can't explain at this late hour.

Hence the light touch as Geoff recommends (after the downsize), as said above, I find 0.3px radius and no more than 100%, often less, works for the subjects I shoot. YMMV.

The shot of the fisherman's hand and reel is a real contrast of texture and it is hardly surprising that what works for one subject type (e.g.metal) doesn't for another (e.g. well focused skin) as Chris says in the last post (cue the bugle).

Cheers,

[MiniChris]

"Also I don't use LAB, but 125%, 1 pixel is going to be noticeable if applied to a downsized image (= overdone, in my view)"

So, curious, what method do you use? I've tried straight unsharp mask at various levels, and I've not been terribly unhappy with LAB, nor with Luminosity where applicable...I think what threw me was comment about blowing them back up by 2x, etc. It simply didn't dawn on me that there is no way to magnify an image and therefore, a moot point (or as one of my students so aptly wrote, a Mute Point) ..tone deaf, I guess.
I will keep at it until I find one which works best on all images...sighhhhhhhh, back to the drawing board.

[Dave Humphries]

"Also I don't use LAB, but 125%, 1 pixel is going to be noticeable if applied to a downsized image (= overdone, in my view)"

So, curious, what method do you use? I've tried straight unsharp mask at various levels, and I've not been terribly unhappy with LAB, nor with Luminosity where applicable...I think what threw me was comment about blowing them back up by 2x, etc. It simply didn't dawn on me that there is no way to magnify an image and therefore, a moot point (or as one of my students so aptly wrote, a Mute Point) ..tone deaf, I guess.
I will keep at it until I find one which works best on all images...sighhhhhhhh, back to the drawing board.

Hi Chris,

I use USM exclusively - I'm happy with the results (no one else complains anymore either ), so I don't feel the need to go to other methods.

I suspect, as with most things PhotoShop, there are many ways to skin an egg (or cook a cat) - forgive my quaint colloquilisms - and several might provide results indistinguishable from each other to a casual observer (e.g. me). Certainly, I don't recall noticing any sharpening issues with your work Chris (when I can see it - I have TinyPic problems).

I've just got into a habit which works for me, and you have one that works for you, so you probably don't need to go back to the drawing board Besides, you must know more about CS5 than I do by now, so I wouldn't knock what you're doing to your own pics, it was just that doing something for a specific purpose, but then subjecting it to a whole different analysis seemed inappropriate, it was indeed a mute /moot point.

One day I may find 'a better way', then I'll be flogging that here

I do also change what I do based on image content, my standard workflow usually has capture, LCE and post downsize output sharpening, all with USM, but I do vary that if the image content warrants it; for example;
a high iso (1600 and above) original requiring a significant crop, I wouldn't do the capture sharpen stage.
a soft original, as well as LCE, I would 'octave' sharpen; several passes of USM; e.g. 3px at 40% + 1.5px at 50% + 0.7px at 60% downsize then 0.3px at whatever is needed 50-70% or so
Sometimes I'll only do the sharpening on the subject itself, if I am fighting 'too much too sharp' problems (insufficient bokeh)
Other times (rare for me) with a subject like the fisherman's hand and reel, I'd do two different methods for the different textures.

In time, I expect I will learn other methods, probably from people like yourself.

Perhaps I should add a line to my signature; "Just because I'm a mod here doesn't mean I'm always right, or even know what I'm doing half the time" because that is exactly how I view myself.