The concepts of autocovariance function and Wiener spectrum have been applied to describe the recorded image noise in nuclear medicine. They were derived as functions of the expected detected count density and the detector and exposure point spread functions. It was shown that the detector system affects only the noise magnitude, whereas the recorder system affects both noise magnitude and texture. In experimental studies, the autocovariance function and Wiener spectrum of recorded image noise were measured by one-dimensional time-series analysis. Due to the non-linearity of the recording film, the best agreement between the theoretical predictions and the experimental results is found when the detected count density is sufficiently high and the size of the exposure spot is sufficiently large for the density fluctuations in the recorded noise image to be relatively low.