Microarray technology has provided an opportunity to simultaneously monitor the expression levels of a large number of genes in response to intentional perturbations. A necessary step towards successful use of microarray technology is background correction which aims to remove noise. One of the most popular algorithms for background correction is the Robust Multichip Average (RMA) procedure which relies on an unjustified parametric assumption. In this paper we first check the fitness of the RMA model using a graphical approach and then propose a new background correction method based on a semiparametric RMA model (semi-RMA). Evaluation of the proposed approach based on spike-in data and MAQC (Microarray Quality Control project) data shows our semi-RMA model provides a better fit to microarray data than other approaches.