In cryo-electron microscopy (cryo-EM), the Wiener filter is the optimal operation – in the least-squares sense – of merging a set of aligned low signal-to-noise ratio (SNR) micrographs to obtain a class average image with higher SNR. However, the condition for the optimal behavior of the Wiener filter is that the signal of interest shows stationary characteristic thoroughly, which cannot always be satisfied. In this paper, we propose substituting the conventional Wiener filter, which encompasses the whole image for denoising, with its local adaptive implementation, which denoises the signal locally. We compare our proposed local adaptive Wiener filter (LA-Wiener filter) with the conventional class averaging method using a simulated dataset and an experimental cryo-EM dataset. The visual and numerical analyses of the results indicate that LA-Wiener filter is superior to the conventional approach in single particle reconstruction (SPR) applications.