What Are Upsampling & Downsampling in Photoshop?by Tony AllevatoUpdated September 28, 2017
Upsampling and downsampling are two specific instances of an imaging process known as "resampling." When you make an image larger or smaller in resolution or dimensions, Adobe Photoshop must resample the image to determine how to fill in the new pixels created between the originals (when enlarging) or which of the original pixels should be retained and which should be discarded (when reducing the size).
When you take a photograph of a natural scene using a digital camera, the camera must convert that scene into a discrete grid of pixels. The same is true for images acquired onto a computer from other physical sources, such as a scanned photograph or document. This process is called "sampling."
Once an image is in a digital medium, you no longer have access to any other information about the scene from which it originated beyond the pixels in the image. This means that when you resize an image, Photoshop cannot fill in gaps in the data with information from the actual scene; instead, it must "resample" the image to construct an approximation to the original scene.
Upsampling is the type of resampling that Photoshop performs when you enlarge an image. When the image is enlarged, the original pixels will then be spaced further apart, and the application must "make up" the new pixels between them, using an appropriate approximation of the original scene.
Downsampling is the opposite of upsampling, used when an image is made smaller. Although shrinking an image does not require filling in new space as in the case of upsampling, Photoshop may still use an approximation in order to preserve as much information about the image as possible.
For example, consider an image made up of alternating black and white pixels. If you shrink this image to half its size by directly sampling the values of every other pixel, you would end up with a completely white or black image. A better approximation method is needed to ensure that such details are not lost.
The approximations to the original scene used by upsampling and downsampling are generated by a process called "interpolation." When you enlarge an image, Photoshop interpolates new pixels in the image by computing averages from the surrounding pixels that it knows about in the original image. It performs similar computations when you shrink an image to determine how to blend the original pixels to generate the new image.
Photoshop supports three main methods of interpolation. Nearest-neighbor interpolation fills each new pixel in an enlarged image with the same color as the closest original pixel to it, but this causes images to appear "blocky" and low in quality.
Bilinear and bicubic interpolation are methods that result in higher-quality images. Bilinear interpolation computes averages of the nearest four surrounding pixels; bicubic uses 16 surrounding pixels. In many cases the images generated by these two methods will look similar, but the extra information used by bicubic interpolation helps to preserve fine detail, both when enlarging and shrinking images.
- Adobe Photoshop CS4: Resampling
- Cambridge in Colour: Understanding Digital Image Interpolation
- "Digital Image Processing, Second Edition"; Rafael C. Gonzalez and Richard E. Woods; 2002
- magnify the needles image by timur1970 from Fotolia.com