Harry Nyquist says resize your damn photos
In yet another chapter of " saying it once to avoid repeating myself" I'd like to discuss the concept of Nyquist Frequency or Nyquist Rate as related to posting or emailing photographs.
The wikipedia entries on Nyquist-Shannon sampling theory are pretty technical, so I'll explain in simple terms. Your digital camera creates a digital sample of the image it sees. It takes the continuous image coming through the lens and breaks it into discrete pixels. The Nyquist Rate tells you the minimum number of pixels required to represent the image: "twice the highest frequency contained within the signal". In other words if you're trying to take a picture of a white picket fence with 100 evenly spaced slats your photo will need to be at least 200 pixels wide - one pixel for the slats, one for the gaps. Anything smaller than that and the details blur into single pixels. That's what's happening in the picture to the right: when the frequency of the detail goes below the nyquist limit you lose that detail.
You need to decide what the highest significant detail is, to set your nyquist limit and the sample size. For example mastering a CD at 44Khz is "good enough" for most people (but not everyone) because the upper range of human hearing is around 22Khz. You also need to decide what's not significant in your image before you (re)sample it. Do you want to see the nails holding the fence together? The individual paint flakes on the fence? The grains of silver in the film you scanned? The noise in the photodetector that scanned the film?
Maybe you want to save all that data in archival offline files, but if I've pointed you to this page it's probably because I don't think you actually thought I needed that much information. Some images need to be very large to represent important detail - maps are a good example - but most of the time huge files are unnecessary. The Nyquist-Shannon sampling theory says that pictures should be just big enough that smallest significant features are one or two pixels large. When in doubt, send the smaller image and wait for someone to ask for the high-res version.
The wikipedia entries on Nyquist-Shannon sampling theory are pretty technical, so I'll explain in simple terms. Your digital camera creates a digital sample of the image it sees. It takes the continuous image coming through the lens and breaks it into discrete pixels. The Nyquist Rate tells you the minimum number of pixels required to represent the image: "twice the highest frequency contained within the signal". In other words if you're trying to take a picture of a white picket fence with 100 evenly spaced slats your photo will need to be at least 200 pixels wide - one pixel for the slats, one for the gaps. Anything smaller than that and the details blur into single pixels. That's what's happening in the picture to the right: when the frequency of the detail goes below the nyquist limit you lose that detail.
You need to decide what the highest significant detail is, to set your nyquist limit and the sample size. For example mastering a CD at 44Khz is "good enough" for most people (but not everyone) because the upper range of human hearing is around 22Khz. You also need to decide what's not significant in your image before you (re)sample it. Do you want to see the nails holding the fence together? The individual paint flakes on the fence? The grains of silver in the film you scanned? The noise in the photodetector that scanned the film?
Maybe you want to save all that data in archival offline files, but if I've pointed you to this page it's probably because I don't think you actually thought I needed that much information. Some images need to be very large to represent important detail - maps are a good example - but most of the time huge files are unnecessary. The Nyquist-Shannon sampling theory says that pictures should be just big enough that smallest significant features are one or two pixels large. When in doubt, send the smaller image and wait for someone to ask for the high-res version.