In this work, we developed RMLPCA, an R package (available on R CRAN) that implements Maximum Likelihood Principal Component Analysis (MLPCA), a method first introduced by Peter D. Wentzell and co-authors in 1997 and used to apply principal component analysis on data with measurement errors. MLPCA uses the principle of maximum likelihood estimation to estimate error-free data given the noisy observations and their measurement error covariance structure. RMLPCA includes algorithms to compute MLPCA-based predictions for different error structures, varying from i.i.d errors to correlated errors within observations and variables. We demonstrate the use of RMLPCA with examples involving dimensionality reduction and image denoising. In the first example, we added heteroscedastic errors on error-free simulated data to show the influence of heteroscedastic errors on PCA and how MLPCA can be used to obtain better results. In the next one, contaminated images are denoised to get a clean version.