Last updated on

23 Apr. 2007

Upper page

Resampling

Image Resampling Algorithms

Image resampling operation changes the size of a digital image. Conventional algorithms include Nearest Neighbor, Bilinear, and Bicubic interpolation. Other linear filters are also known: Lanczos, Mitchell, Catmull-Rom, and other splines. Here we compare visual quality of these resampling filters, and also review several image-adaptive edge-directional approaches to image resampling. Also we introduce a new method for image resampling, we call it SmartEdge.

SmartEdge algorithm was developed by the Graphics and Media Lab of a Moscow State University (MSU) in a joint research project with Samsung Electronics.

SmartEdge 2 algorithm is a newer version of SmartEdge algorithm developed for top-notch quality upscaling. Notice the increased level of details and fine texture in the images processed with this algorithm.

Demo program download

Demo program for Windows that implements SmartEdge 2 algorithm is available for free non-commercial use. Please note that the algorithm is currently very slow and consumes large amounts of memory, so it's only applicable to small images.

Download SmartEdge 2 demo application (ZIP)

Example

At the enlarged images we show only image fragments: click on them to view full enlarged images. Some of the images on this page are losslessly encoded in PNG format, but some larger images were encoded in JPEG format (at very high quality).

Original low-resolution image - it will be enlarged by 2 times

Hover with a mouse over the algorithm name to preview the result.
Click on the algorithm name to get the full image.
SmartEdge 2 result is displayed by default.

Other methods...

Another example

Original low-resolution image - it will be enlarged by 2 times

Up-scaled images

Hover with a mouse over the algorithm name to preview the result.
Click on the algorithm name to get the full image.
SmartEdge 2 result is displayed by default.

Artifacts

Let's discuss the characteristic artifacts introduced by resampling methods.

Linear methods

Linear resampling methods are always a tradeoff between 3 artifacts: blur, aliasing, and ringing.

It's important to understand that it's theoretically impossible to eliminate all of these artifacts simultaneously with linear methods.

Edge-adaptive methods

The most widely known edge-adaptive resampling method is NEDI - New Edge Directed Interpolation - introduced by Xin Lee. It has several artifacts.

Random pixels appear sometimes in smooth image areas or near smooth image areas.
Pattern areas can have large artifacts. They become visible in areas of repeating patterns of small size.
Areas of fine structure can look unnatural. This includes many natural textures, such as leaves, grass, clouds, water.