Today’s Question: How much can you safely upsize an image without sacrificing noticeable quality degradation? And what is the best resampling method for upsizing? Bilinear, Bicubic, or Lanczos?
Tim’s Quick Answer: As a very general rule you can enlarge a photo to double the horizontal and vertical dimensions (quadruple the total area), though the specifics depend on how the image will be shared. I recommend the Bicubic interpolation algorithm in most cases.
More Detail: There really isn’t any strict rule about how large you can print a photo while maintaining good quality. As I often explain when it comes to this topic, you can make a photo infinitely large as long as the viewer will be extremely far away from the image. If you don’t get too close to an extreme enlargement, you won’t be able to perceive the degraded image quality.
In most cases, however, I assume that when a photographer wants to know how much they can safely enlarge a photo it is with the assumption that the viewer will be able to get relatively close to the enlarged image.
In my experience as long as the source image is of good quality in terms of sharpness, detail, and minimal noise, you can comfortably enlarge a photo to double the height and width (quadruple the total image area) while maintaining good image quality. That’s not to say you can’t produce great results with an even bigger enlargement, but doubling the dimensions will generally ensure good image quality in the enlargement.
When it comes to the interpolation algorithm used for enlargement, in general I recommend the bicubic algorithm. For significant enlargements Photoshop includes a “Bicubic Smoother” algorithm, which can be advantages for especially significant enlargements.
The Lanczos algorithm is also considered to be very good, though it is most commonly used for enlarging video rather than still photos. There is a risk of contrast artifacts along edges within a photo when using Lanczos, which is why I tend to favor Bicubic.
There are also third-party tools such as Gigapixel AI from Topaz Labs that use methods that go beyond simple algorithms, which can also improve results when creating significant enlargements of a photo.