Downsampling problems

hi

most of time i shot L quality with my 20D, but I destin the pictures quite only for the web in 800X600 size, so I have to reduce hardly the size from 3500 to 800 large, and I notice a sensible loss of quality with this downsampling, while respecting the rules and photosop procedure.

perhaps I made an error shooting in L, may be better shooting directly in M or inferior quality will be better for me, having a least minor size reduction to do ?
or perhaps in RAW ? I don't know...
--
http://pbase.com/isogood
http://isogood.blogphotography.com

...or other software designed to resize the photos with no loss or resolution?

Ben

and I don't see the difference on my monitor. The pictures I see, for the most part, look just fine, and a lot of these pictures where shot in RAW.

Hi, i,

isogood said:

hi
most of time i shot L quality with my 20D, but I destin the
pictures quite only for the web in 800X600 size, so I have to
reduce hardly the size from 3500 to 800 large, and I notice a
sensible loss of quality with this downsampling. . .

Click to expand...

Of course. That's why we would rather have higher resulution when we can. When we have to give it up for file size reasons, we lose that benefit.

Your downsized images will have less than 1/4 teh resolution of the originals.

Best regards,

Doug

Visit The Pumpkin, a library of my technical articles on photography, optics, and other topics:

http://doug.kerr.home.att.net/pumpkin

'Make everything as simple as possible, but no simpler.'

At full size I find bayer images just a tad soft, after downsizing they are as packed with detail as possible for their size. Though naturally a 3500 pixel wide image is going to have much more detail than a 800 pixel wide one. No technique in the world will change that.

Though you need to apply a light sharpening to make them snap. Use a very small sharpening radius (.1 or .2) and increase the level till you are happy.

Any proper resize algorithm is going to introduce some softness that requires some USM, but on downsizing, it usually looks great with a very light touch.

isogood said:

hi
most of time i shot L quality with my 20D, but I destin the
pictures quite only for the web in 800X600 size, so I have to
reduce hardly the size from 3500 to 800 large, and I notice a
sensible loss of quality with this downsampling, while respecting
the rules and photosop procedure.
perhaps I made an error shooting in L, may be better shooting
directly in M or inferior quality will be better for me, having a
least minor size reduction to do ?
or perhaps in RAW ? I don't know...
--
http://pbase.com/isogood
http://isogood.blogphotography.com

Click to expand...

--
I see dead pixels

Regards

Thanks for answers

I have tried fred Miranda, with soft sharpening before downsample, but not better than Photoshop (or i use it wrong way ...)
I will try genuine fractal.
--
http://pbase.com/isogood
http://isogood.blogphotography.com

I always use the unsharp mask after resampling.

--
Todd Muskopf
professional fine art painter, aspiring photog
http://www.muskopf.org

Post one unimportant image full size and we can try different things. This didn't seem to have the detail I was expecting, but I gave it just enough USM to clear the softness. You could use some resample method that does auto sharpening, but you may prefer the control of sharpening afterward.

isogood said:

hi
most of time i shot L quality with my 20D, but I destin the
pictures quite only for the web in 800X600 size, so I have to
reduce hardly the size from 3500 to 800 large, and I notice a
sensible loss of quality with this downsampling, while respecting
the rules and photosop procedure.
perhaps I made an error shooting in L, may be better shooting
directly in M or inferior quality will be better for me, having a
least minor size reduction to do ?
or perhaps in RAW ? I don't know...
--
http://pbase.com/isogood
http://isogood.blogphotography.com

Click to expand...

Sharpening for the final, intended use of the photo should always be the final step. From my (admittedly pedestrian) experience, and from what I've read, you should always reserve 'room' for this step at the end... in other words, don't be overly agressive in sharpening until you get to your final output (prepping for print, web, wallpaper, whatever). This is a very simplistic expression of what is contained in this very informative article:

http://www.luminous-landscape.com/reviews/software/pk-sharpener.shtml

While it's centered on a review of a specific product, it embeds some very sound principles that can be adhered to, regardless of the tools being used to sharpen an image.

It's a good read and especially revealing if you've never "oversharpened" (in an onscreen visual sense) prior to printing. The results are exceptional, once you get past the nauseating on-screen look of what you're about to ask your printer to digest. =D

Bottom line: Don't expect to simply downsample your great looking, sharp image and expect it to retain it's sharpness. You should be sharpening for final output.

icmp

All the replies here are sensible, but they assume you mean different things. What are you refering to? Do you want the best image that will fit in 800 pixels as viewed at full size?

If you make your image fit on your monitor, you are resampling. You can't compare the full detail of the original, where only a corner can be viewed at a time, with one that fits fully on the screen. If the intention is to resample an image to make it fit on a screen, say at 800x600, most resampling approaches will give a better image than just "fit to screen" in PS or other software.

There are many ways to get a smaller image. I have FM's plugin, Genuine Fractals, and I have tried bicubic, etc, etc. None of them do very much, to be honest. The key is to sensibly resharpen after resizing. If you do that, then a simple resize plus sharpenning will usually match any of the more complicated approaches. In that respect, the FM plugin is a ripoff.
--
Charlie

Better lucky than smart. The more shots I take, the luckier I get.

... and to me its one of the most compelling arguments for having more MPixels.

Yep, you always need to sharpen after downsampling an image. The reason is that you are effectively throwing away a LOT of pixels, and as you do so, the edge pixels between light and dark areas will be antialiased. You have less pixels to describe the same image, so your software often needs to make a guess about what color value a pixel should have.

Sharpening will increase the contrast on the edge-pixels again and your image should "snap" once more. Unsharp mask is good, my favorite is KPT Equalizer's "Bounded sharpen" from the KPT 6 collection.

--
peace,
Tormod in Stockholm
http://www.airwhale.com/

Thanks for all answers, very instructive, I agree most of them, but I still don't understand why Fred Miranda's tools advertising windows hardly recommend to do the sharpening before downsampling ?

I tried before, I tried after, can't see great difference, but still for use only on the web, not for printing.
Perhaps the workflow recommandation of Fred is only for printing output ?
Better ask him...

Finally, for the web, the target is to keep a maximum of the details at size 800 large, OK, maximum as possible.
--
http://pbase.com/isogood
http://isogood.blogphotography.com