MOTIVATION: Existing microarray genotype-calling algorithms adopt either SNP-by-SNP (SNP-wise) or sample-by-sample (sample-wise) approaches to calling. We have developed a novel genotype-calling algorithm for the Illumina platform, optiCall, that uses both SNP-wise and sample-wise calling to more accurately ascertain genotypes at rare, low-frequency and common variants. RESULTS: Using data from 4537 individuals from the 1958 British Birth Cohort genotyped on the Immunochip, we estimate the proportion of SNPs lost to downstream analysis due to false quality control failures, and rare variants misclassified as monomorphic, is only 1.38% with optiCall, in comparison to 3.87, 7.85 and 4.09% for Illuminus, GenoSNP and GenCall, respectively. We show that optiCall accurately captures rare variants and can correctly account for SNPs where probe intensity clouds are shifted from their expected positions. AVAILABILITY AND IMPLEMENTATION: optiCall is implemented in C++ for use on UNIX operating systems and is available for download at http://www.sanger.ac.uk/resources/software/opticall/.