File:Richardson-Lucy deconvolution gone wild.gif

English: A demonstration of the Richardson–Lucy deconvolution algorithm to some synthetic one-dimensional data.

The top left plot shows the unknown signal (sometimes called the source). The orange region is the true original signal, and the red curve is the signal inferred by applying some number of Richardson–Lucy iterations to the measurement.

The top right plot shows the convolution function (called the impulse-response function in 1D or the point-spread function or kernel in 2D) that describes the transformation from the signal to the measurement.

The bottom plot shows the measured degraded signal (sometimes called the image). The orange region is the actual measurement, with some Poisson noise, as is realistic for many forms of measurement. The red curve is the convolution of the inferred signal with the impulse response function.

The plot stops after 7000 iterations, which is sufficient for Richardson–Lucy to overfit the noisy data considerably. For a shorter animation that stops when the fit is still reasonable, see File:Richardson-Lucy deconvolution.gif.

This image was generated using the code available at github.com/jkunimune/richardson-lucy.