Seem to lose sharpness of an image when it is reduced -why?

Hello,

I took a few images at a photo shoot & when the images are about 3200 x 2000 they look nice & sharp.

I was wondering why when reduced to 800 x 600 for putting on a website they seem to lose a degree of sharpness.

I converted the images from RAW to jpeg,should I be saving the images in a different format?

Advice appreciated please.

cookey,
If you start with 3200 pixels and reduce to 800 pixels where did the 2400 other pixels go? That maybe the reason why you usually have to apply USM or sharpening to a reduced sized file to get it to equal sharpness as the original.

I did the sharpening before I resized my picture,so it is best not to do any PP before resizing then & then do it after?

This is a consequence of resizing from a larger image to a smaller one and happens with all image types and programs used to resize (to differing extents).

Think about it; the program you use to resize the image has to take the 6.4 million pixels in your original and represent this in 480k pixels. To do this the program uses all sorts of clever interpolation algorithms, along the way trying to maintain sharpness, contrast, avoid moiré patterns etc, etc.

The moral of the story is to apply a little sharpening after you have resized (generally USM or high pass is best). Most folks include the sharpening as part of their resize batch processing.

thanks very much for the tips

If you use PS, give Manyk SRS a try.

Also, (assuming you're using Photoshop) when you resize the image make sure you choose Bicubic Sharpen instead of plain old Bicubic.

This might be a little off topic, but does anybody know if Lightroom applies sharpening automatically to files when exporting a RAW file to a small jpeg, say 800x600? I swear some of my pictures look sharper as a jpeg export than they did in the Develop mode in Lightroom.

guitarman2977 wrote in post #5394485
This might be a little off topic, but does anybody know if Lightroom applies sharpening automatically to files when exporting a RAW file to a small jpeg, say 800x600? I swear some of my pictures look sharper as a jpeg export than they did in the Develop mode in Lightroom.

I can't answer your question about LR sharpening on output, but I have noticed that images look sharper in LR when viewed at 1:2, 1:3, 1:4 sizes rather than Fit or Fill. Whenever I resize anything for the web I always try to resize to 1:integer rather than some random reduction ratio just to fit an exact pixel size.

e.g. files from my 40D are 3888 * xxxx pixels. I always output to 972 * xxx (1:4 ratio) or 778 * xxx (1:5 ratio). If I aim for 1024 *xxx or 800 * xxx, for example, the final images do not look as good to me. The same thing goes for files viewed within or output from DPP too.

tdodd wrote in post #5394608
I can't answer your question about LR sharpening on output, but I have noticed that images look sharper in LR when viewed at 1:2, 1:3, 1:4 sizes rather than Fit or Fill. Whenever I resize anything for the web I always try to resize to 1:integer rather than some random reduction ratio just to fit an exact pixel size.

e.g. files from my 40D are 3888 * xxxx pixels. I always output to 972 * xxx (1:4 ratio) or 778 * xxx (1:5 ratio). If I aim for 1024 *xxx or 800 * xxx, for example, the final images do not look as good to me. The same thing goes for files viewed within or output from DPP too.

That's a great idea. Thanks! I'll have to try exporting to pixel dimensions that are a factor of the original. Does anyone else on here resize this way?

What I usually do is apply USM to the large image with a radius of 1 pixel, and threshold somewhere between 1-5.
Then resize, and apply smart sharpen to the resized picture around 35% and Gaussian blur removal mode. When I resize, I try to resize to an even multiple of the original size. IE, if your image is 4000 pixel, resize it to either 2000, 1000, 400, etc pixels. When you view the resized image, make sure you select the "View Actual Size" option as the interpolation LCD displays do on off sized images will often times make it look far worse than the actual image.

bsmotril wrote in post #5394749
What I usually do is apply USM to the large image with a radius of 1 pixel, and threshold somewhere between 1-5.
Then resize, and apply smart sharpen to the resized picture around 35% and Gaussian blur removal mode. When I resize, I try to resize to an even multiple of the original size. IE, if your image is 4000 pixel, resize it to either 2000, 1000, 400, etc pixels. When you view the resized image, make sure you select the "View Actual Size" option as the interpolation LCD displays do on off sized images will often times make it look far worse than the actual image.

So if I crop a RAW image in post processing, do I need to figure out the new pixel width and export to a factor of that number, or does it work to still base it off the original image size?

guitarman2977 wrote in post #5394485
This might be a little off topic, but does anybody know if Lightroom applies sharpening automatically to files when exporting a RAW file to a small jpeg, say 800x600? I swear some of my pictures look sharper as a jpeg export than they did in the Develop mode in Lightroom.

AFAIK LR 1.4 doesn't offer output sharpening LR 2.0 beta does.

guitarman2977 wrote in post #5394970
So if I crop a RAW image in post processing, do I need to figure out the new pixel width and export to a factor of that number, or does it work to still base it off the original image size?

You'd have to calculate the output dimensions based on the cropped image, not the original. Once you've cropped to a new size, it doesn't matter at all what size the file started out at. You've effectively thrown some pixels away. Pretend they were never there. The idea is that....

for a 1:2 ratio you are squeezing exactly 2X2 input pixels into one output pixel;
for a 1:3 ratio you are squeezing exactly 3X3 input pixels into one output pixel;
for a 1:4 ratio you are squeezing exactly 4X4 input pixels into one output pixel

This means that the software is not having to guess at what, for argument's sake, squashing 2.36X2.36 or 3.437X3.437 pixels into one should look like. How can the software figure out what those odd bits of pixels should look like - it has to guess. When you squash by integer factors it keeps the maths clean and simple and the software can do a cleaner job of calculating what the new, shrunken pixels should look like.

Well that's my theory

Great information, tdodd. That makes perfect sense. Thanks!