Abstract Motivation There is a growing need for efficient imputation methods for high-dimensional DNA methylation (DNAm) datasets. Existing microarray imputation approaches, such as k-nearest neighbors (K-NN) or principal component analysis (PCA)-based methods, provide high accuracy but can be computationally intensive, while methods for whole-genome data are not designed for large cohorts. We developed slideimp, an R package that implements sliding window, groupable, parallelized K-NN and optimized PCA imputation to address these limitations. Results Benchmarks on microarray DNAm datasets demonstrate that slideimp achieves up to 150x faster runtime and 10x-100x lower memory usage while maintaining comparable or superior accuracy over existing methods and implementations. For K-NN and PCA imputation, slideimp supports grouped imputation which enhances imputation efficiency and accuracy. In a whole-genome dataset, sliding window K-NN imputation substantially increased the correlation of the Horvath 2013 clock with chronological age from 0.131 to 0.477. Additional features include targeted imputation of CpG subsets for K-NN and estimation of imputation accuracy via repeated cross-validation. The efficient and flexible DNAm imputation methods implemented by slideimp can easily be applied to other high-dimensional data types. Availability and Implementation The code to fully reproduce all analyses presented in this paper is available on GitHub at https://github.com/hhp94/slideimp_paper . The R package slideimp is also available on GitHub at https://github.com/hhp94/slideimp . Supplemental figures are available online.