What’s all the NOISE about noise? - Part Two

In the first article in this series, I discussed why judging the image quality of a camera, based on a small jpg image posted on a photography forum is a bad idea. The reason being that downsizing an image hides flaws. In this article I will discuss the flip side of this argument, or why basing your judgment of image quality(eg. sharpness, NOISE) solely on viewing the files at 100% can be also be misleading.

To be honest, viewing an image at 100% magnification(where one image file pixel equals one monitor pixel) is the only way to judge true image sharpness and quality. But, it can also lead you astray. While it certainly discloses any flaws in the image, it doesn’t  tell you how it will look printed at a given size. When an image is printed commercially for billboards or posters, the anticipated viewing distance must be taken in to account. There is no point in printing at a high resolution if the viewer will be too far away to resolve the extra detail. A typical 8x10 would be printed at 300dpi(although not quite the same, you can think of it as 300 pixels per inch). At an arm’s length viewing distance, this gives the print sufficient resolution. A billboard on the other hand, may only be printed at 10 - 20 dpi because it is expected that the viewer will be some distance away. When viewed at a distance, we perceive the billboard as “sharp”, however if we walked up to 3 or 4 feet away, we would see that there is little detail and it is “blurry”.

A general rule of thumb that printers use in determining viewing distance is to multiply the diagonal of the print by 1.5 or even 2 times. For a 4x6 inch print, this gives 11 “ (at 1.5x) as the expected viewing distance. The dpi required to give us detail and sharpness is around 300dpi, which is the typical printing resolution for a 4x6.

So what if we take this information and apply it to your image viewed at 100% on your monitor? Let’s use the 21 megapixel Canon 5Dmark2 as an example. The files are 5616 pixels wide by 3744 pixels high. A decent monitor will display around 90 pixels per inch. If we want to see how that correlates to a print, divide the image pixels by 90(5616/90 and 3744/90). It gives dimensions of 62 inches by 41 inches. This means when we are looking at our 5Dm2 files at 100% , from a viewing distance of 16-20 inches, it is equivalent to a print over 5 feet wide!

If we now take our viewing formula(1.5 or 2x the diagonal) and apply it to our example, we find that a print 62” x 41” would be expected to be viewed from a distance of over 12 feet away and not typical monitor viewing distances of 16 - 20 inches(62x41 gives a diagonal of 74”, 2x 74 “= 148” or 12.3 feet)! What we are doing by viewing at 100% is akin to walking up to a billboard and examining it from a few feet away and expecting it to have the detail of a 4x6 print!

In summary, while it is necessary to examine a file at 100% to determine true image quality, it does tend to exaggerate flaws. So before you start tossing out your images due to “noise” or sharpness issues, think carefully about the intended end use, the image you send to the trash bin might be just fine for your purposes!

This is the jpg that I posted in Part One of this series. Below is a 100% crop of a section of the wing of the Whitebreasted Nuthatch, taken at ISO 102,400. The crop section illustrates two points. The first point shows just how much noise has been hidden in the downsizing. The second point is that noise + detail is a much better outcome than LOW noise + little detail. This image is noisy, but it has detail, and if you have detail,  then you have latitude in your output options whether it be printing or web output.

Technical notes: The crop section is a pixel by pixel crop of the original, NO noise reduction has been applied, no sharpening, other than defaults in Aperture.