Sharpening and Downsizing for Web

[fike]

I have been battling with sizing and sharpening my online portfolio. I typically do very large and detailed panoramic and mosaics.  My smallest images that I sell are printed 12.5"x30" at 240 ppi.  When I downsize to an 800 pixel wide image and then sharpen, I am rarely satisfied with the results.  Pebbles and stones, leaves and grass, twigs and branches, whatever detail there is seems lost in the small image.  These are my compositional elements.  I want to preserve the little stuff, but I sometimes think it is beyond the online medium to do a satisfactory job.

I know there will be loss of detail when downsizing and compressing, but what do you do to maintain detail on your detailed images online?

My process has been

1)Open image in CS3 and flatten
2)Resize to 800 pixels wide x N pixels tall using Bicubic Sharper
3)Use Smart Sharpen with remove Lens Blur selected.  Generally Radius between 0.2 and 0.6 and amount of 80%-150%.
4)Add border and title.
5)Save for Web and target between 100K and 200K for large images.

Results have been better than in the past, but still not excellent, in my opinion.  Check http://marcshaffer.net for examples.

Are there any plugins that might help me do a better job.  I am interested in focusing on printing my images, not optimizing for web viewing.  Shortcuts are just fine with me?

[fike]

As I scan further down the topic, I realize what a hot topic this is.  To clarify, I am not concerned with the sizing or color profiles aspect as much as maximizing detailed images with sharpening before compressing to jpg.  Resizing is a relevant part of sharpening, but I do understand the sizing issue well enough.

thanks, sorry to duplicate topics.

[Wayne Fox]

I struggle with this as well, most images are from a P45, or even multiple P45 captures stitched.

Lately I've been using similar steps to yours, but I have added a very slight smart sharpen before downsizing. I also always use at least a slight capture sharpening when processing the raw file.  This seems to help hold some of the clarity.

[Tim Gray]

It's probably just my imagination, but using the LR web gallery vs what I used before (the Russell Brown script that resized, converted to sRGB etc. etc. ) my web stuff now seems acceptably sharp (or at least sharper).  I was never able to consistently land on a web output sharpening technique that I was happy with (including the Photokit tool).  

Compare, fi,

http://www.timgrayphotography.com/gallerie...misc/index.html  (very so so)

to

http://www.timgrayphotography.com/gallerie...of_t/index.html  (seems a bit better - this is my first LR gallery)

Both were processed (pre web) with similar techniques, neither has specific output sharpening applied. At least I don't think I'm sharpening as part of the LR Gallery process

[fike]

I struggle with this as well, most images are from a P45, or even multiple P45 captures stitched.

Lately I've been using similar steps to yours, but I have added a very slight smart sharpen before downsizing. I also always use at least a slight capture sharpening when processing the raw file.  This seems to help hold some of the clarity.
[a href=\"index.php?act=findpost&pid=155024\"][{POST_SNAPBACK}][/a]

I have read about doing this too, but my experiments with this have been inconsistent.  The difference between sharpening a 3000x2000 pixel frame are very different from a 12,000 x 5,000 pixel image, especially when your destination is 800x400 pixels.  

I have considered doing the rsize and sharpening in a few steps, but that is very time consuming.

[fike]

It's probably just my imagination, but using the LR web gallery vs what I used before (the Russell Brown script that resized, converted to sRGB etc. etc. ) my web stuff now seems acceptably sharp (or at least sharper).  I was never able to consistently land on a web output sharpening technique that I was happy with (including the Photokit tool). 

Compare, fi,

http://www.timgrayphotography.com/gallerie...misc/index.html  (very so so)

to

http://www.timgrayphotography.com/gallerie...of_t/index.html  (seems a bit better - this is my first LR gallery)

Both were processed (pre web) with similar techniques, neither has specific output sharpening applied. At least I don't think I'm sharpening as part of the LR Gallery process
[a href=\"index.php?act=findpost&pid=155037\"][{POST_SNAPBACK}][/a]

I also had poor results with the Russel Brown automation.  It is very convenient, but the output isn't that great.  I don't think you can apply a one-size fits all aproach to sharpening.  

Looking at the two galleries, I agree that the former is not as sharp as the latter, but I also think that subject matter enters into it.  Buildings with straight lines accept sharpening much easier than irregular leaves and bark and stuff like what you have in the first set of images.  

(While I agree that the sharpening on the first set isn't as good, there are some fantastic skies and tree silhouettes in that set.)

When sharpening for print, I always view at 100% and then adjust the sharpness until I barely notice the change, then I back off.  When sharpening for the web, I generally adjust until I bareley notice the change and then go just beyone--with the consideration that some sharpness will be lost in the compression.  I wonder if that is the right approach.  

I am also considering applying sharpening masks to sharpen leaves and trees with different settings than foreground details and sky.  It seems like a pain, but I haven't yet come up with any settings that I think are excellent for leaves and trees anyway.

[Mark D Segal]

Are there any plugins that might help me do a better job.  I am interested in focusing on printing my images, not optimizing for web viewing.  Shortcuts are just fine with me?
[a href=\"index.php?act=findpost&pid=155012\"][{POST_SNAPBACK}][/a]

I use PK Sharpener. It has customized settings for a host of printing conditions as well as web output. On the whole I'm very pleased with it. I've described and discussed all this further in two other neighbouring and active threads very recently on preparing images for the web.

[Tim Gray]

I'm basically reconciled to not customizing sharpening for specific web posted shots.  As I indicated above, I'm hoping the LR Web functionality will help things a bit.  I judge the quality of my images by prints, not web and to me it seems that any effort directed at a specific 800 x 600 image is almost an exercise in futility.  Having said that, I don't sell via the web and if there was a monetary incentive to spend additional time on each posted image, I'd reconsider.

[jbrembat]

When you downsample an image you can't retain all the details.
You can try, if you like, an  increase of the global contrast and  of the local contrast.



jacopo

[fike]

When you downsample an image you can't retain all the details.
You can try, if you like, an  increase of the global contrast and  of the local contrast.

Hmmm....that image is a very good example of my problem.  The composition depends upon detail as much as anything.  I agree that I can't keep all the detail, but how do I throw it away gracefully.  

The modification you made looks good, but I can't decide if I prefer it.  What tools did you use for local and global contrast?  I may try it on the full-sized image before sizing and sharpening.

For others to see, here is my original version before JB made modifications.

[jbrembat]

What tools did you use for local and global contrast?

Global contrast was made with an S-shaped curve on the luminosity.
Local contrast was made with USM applied on the luminosity.

I used my ow PhotoResampling product.

Jacopo

[Jonathan Wienke]

I generally resize to web dimensions with standard Bicubic, then Focus Magic radius 1, strength 25-50%, depending on how crispy the image needs to look. This image is strength 50%:

[attachment=4008:attachment]

IMO Focus Magic does a good job of maximizing perceived detail without introducing haloes or other undesirable artifacts.

FWIW, here is Fike's image run through Focus Magic at 50% faded to 67% (50% was too much, 25% not quite enough, so I split the difference)

[attachment=4012:attachment]

[mistybreeze]

For what it's worth, I prefer the "original version before JB made modifications."

The image in Post #9 looks awfully crunchy to me. I find very little beauty and elegance in overly sharpened digital images.

On my NEC screen, the original version (in Post #10) is far more natural and realistic. I also prefer the color of the moss, which changed dramatically in the modified version.

One really needs to study samples of too much versus not enough to develop a keener sense of how far one wishes to go. The choices are complicated because they're completely subjective. But when they're illustrated in threads like this, it helps.

Thanks for posting the photo.

[Mark D Segal]

I agree with Mistybreeze on all of the above. More generally, one wants a sharpening workflow which makes minimal changes to the desired luminosity and colours, while not over-sharpening the images to the point of making them crinkly. I've found on a number of my images it was better to create the web versions with no output sharpening, while mosts of the others benefit from a very moderate amount.

[fike]

I agree with Mistybreeze on all of the above. More generally, one wants a sharpening workflow which makes minimal changes to the desired luminosity and colours, while not over-sharpening the images to the point of making them crinkly. I've found on a number of my images it was better to create the web versions with no output sharpening, while mosts of the others benefit from a very moderate amount.
[a href=\"index.php?act=findpost&pid=155506\"][{POST_SNAPBACK}][/a]

It's a tough balance.  To a landscape photographer, throwing away detail is sacreligious.

[Mark D Segal]

It's a tough balance.  To a landscape photographer, throwing away detail is sacreligious.
[a href=\"index.php?act=findpost&pid=155626\"][{POST_SNAPBACK}][/a]

Well, it's not so much a matter of throwing away detail - it's the extent to which you emphasize it by increasing edge contrast. I would say the balance is between how fuzzy versus how pointed you want the acutance to be, and that would vary from image to image. With a layer-based approach to sharpening, you can adjust back and forth till you get what you want.

[jbrembat]

I may try it on the full-sized image before sizing and sharpening.

If you like the original large image, don't touch it.
Instead make a web version.

In any case, as downsampling generate aliasing artifacts, a slight blur of the original image can reduce them.
So sharpening before resampling is not a good idea.

My editing may appear a bit strong, but it is intentional. I try to give some editing direction not a final image. So the edting was performed to give a clear difference before/after.

Jacopo