downsamples

Downsampling is the process of reducing the sampling rate or resolution of a signal or dataset by retaining a subset of samples or pixels. In digital signal processing it is often called decimation and is the counterpart to upsampling.

In time-domain signals, downsampling by a factor M retains every M-th sample, yielding a new sampling rate Fs' = Fs/M. Before downsampling, an anti-aliasing low-pass filter is typically applied to limit content below the new Nyquist frequency.

In image processing, downsampling reduces spatial resolution by resampling with interpolation or by pooling pixels. To minimize artifacts, pre-filtering with a low-pass filter or using area-based methods is common; various interpolations balance quality and speed.

In data analysis, downsampling reduces data size for efficiency or to match observation frequency. Methods include aggregating over fixed windows (mean, sum) or selecting regular samples. It can introduce bias if temporal or spatial structure is not preserved.

Practical considerations include choosing the downsampling factor, maintaining information content, and handling borders in images or streaming data. Reconstruction from downsampled data may require upsampling and interpolation.

See also: resampling, decimation, interpolation, upsampling, anti-aliasing.