Optimal resolution/compression when resizing for upload?

[rick7]

Am using IrfanView to batch resize/resample before uploading images.  The excellent Coppermine tutorial about this suggests resizing by setting "long size" to 640 pixels (presumably the width then usually gets set to 480).  My question: why this value?  What are some of the factors involved in deciding what size to resize to?  What are the consequences for your uploaded image of these different settings?  Thanks much.

[kegobeer]

640x480 is a "web friendly" resolution.  Images of this size tend to have smallish file sizes and display well on most people's browsers.

Larger images have longer load times, require more resources to do the resizing and image creation, etc.  It's a matter of preference.  I prefer to keep my visitors happy by keeping images fairly small.  That also keeps my host happy by not sucking all the resources up.

[rick7]

My own monitor is at 800x600 and I guess I'm thinking that if I want to see a crisp full-screen slideshow without jaggies it's best to set the resizing to 800x600.  Am I wrong?  For most people with even a larger resolution, wouldn't 640x480 produce an image without significant jaggies?

[kegobeer]

"jaggies"?

[rick7]

Sorry -- here's a description of 'jaggies' from the Glossary at Digital Photography Review.  It's basically the jagged edges of diagonal and other lines that are one of the primary indicators of poor (or mismatched) resolution:

http://www.dpreview.com/learn/?/Glossary/Digital_Imaging/Jaggies_01.htm

Jaggies
By Vincent Bockaert
   
Hardly a technical term, jaggies refer to the visible "steps" of diagonal lines or edges in a digital image. Also referred to as "aliasing", these steps are simply a consequence of the regular, square layout of a pixel.
 
Increasing Resolution Reduces the Visibility of Jaggies
Jaggies become less visible as the sensor or image resolution increases. The crops below are from pictures of a flower against a blue sky taken with digital cameras with different resolutions*. The low resolution cameras show very visible jaggies. As we increase the camera resolution from A to D, the steps become almost invisible in crop D. But they are still present when the image is enlarged, as shown in crop E.

[kegobeer]

In my experience, resizing from a large image to a smaller image results in pretty good quality, depending on the software used, the various quality settings, etc.  The XP resize spits out poor quality resized images IMO; Photoshop produces great quality.  Sizing up from a small image to a large image almost always produces a terrible quality image, simply because of the guesswork that the software has to do in order to enlarge the image.  Obviously, small to big isn't the issue here.

Most people have limited space available and have to compromise some image quality in order to display the most images possible.  640x480 and 800x600 are still good quality images if resized from an original 1600x1200 or higher resolution image.  The images I have on my website are mostly 640x480, resized in Photoshop from an original of 1600x1200.  More than good enough for my family and friends that visit, and they can print out 4x6 photos that have pretty good quality.  Not as good as the original, but so far no one has complained.

When I resize, I don't limit myself to width.  If the image is taller than it is wide, then I resize 480x640.

[Joachim Müller]

If you have plenty of webspace, I recommend batch-resizing everything to 1024 (long side) on the client (before upload) and then set "Max width or height of an intermediate picture/video" in coppermine's config to 640. This way, both your site visitors will be happy (for the sake of different screen resolutions, loading time etc.) and users who are interessted in large res pics (e.g. for printing) can get them as well by clicking on the intermediate pics.
The settings 640x480 in the IrfanView tutorial was chosen deliberately (as an example); of course you're free to use any size. The tutorial is only meant to give you an idea how to use the batch-resize function in the first place.

Joachim