Resampling

Resampling is the digital process of changing the sample rate or dimensions of digital imagery or audio by temporally or areally analysing and sampling the original data.

Audio

Audio resampling is also called sample rate conversion. This operation in digital signal processing involves converting a sampled signal from one sampling frequency to another. For instance, the output waveform of a digital audio workstation that was processed at 96 kHz must be resampled to 44.1 kHz to be placed on a Compact Disc. The article Sample rate conversion explains how this is done.

Bitmap

A digital image is typically stored as a bitmap, it being a literal map of which pixels are what value, to construct an image. (This is not to be confused with the BMP image file format, which is a method of storing bitmaps in file data. PNG, JPEG and GIF are other equally valid methods of storing bitmaps.)

A bitmap is said to be sampled on each pixel, rather than being supersampled (more than one point of data per pixel) or subsampled (less than one point of data per pixel). Resampling this bitmap involves creating a sample grid, which is overlaid on the pixels. According to how far each grid point is away from the original centre of each pixel, and according to whatever resampling algorithm is in use, the new sample point is given a colour value.

The mathematics behind bitmap resampling is multivariate interpolation in two spatial variables, and done separately for each color channel, which outlines the methods available to bitmap resampling. The simplest method is known as "nearest-neighbour" or "point sampling". The closest pixel center to each sample grid point is used, with no input from other surrounding pixels.

Bilinear interpolation is slightly better than nearest neighbour, where a sample point takes the four closest pixel centers and linearly interpolates their color values according to their distance from the sample point. This method is particularly useful when an image is being enlarged, or transformed or distorted without decrease in average size. Nearest neighbour or bilinear should only be used when interpolation speed is critical, if not, the methods Lanczos resampling or bicubic interpolation are better alternatives.

Supersampling is a method where several sample points are calculated and the results averaged (or combined according to a convolution kernel) to yield the required value on the sample grid. This method is particularly useful when an image is being reduced in size, or transformed or distorted with a decrease in average size.

ee also

* Downsampling
* Upsampling
* Oversampling
* Sampling (information theory)
* Signal (information theory)
* Data conversion
* Interpolation
* Multivariate interpolation
* Subsampling

External links

* [http://www.cambridgeincolour.com/tutorials/image-interpolation.htm Image interpolation] , from cambridgeincolour.com