Copy number variation (CNV) is a major class of variations in the human genome, which has been associated with a wide spectrum of human diseases such as cancer, schizophrenia and autoimmune diseases. In recent years, the advancement of high-throughput sequencing (HTS) technologies has provided an opportunity for CNV detection with unprecedented resolution. Based on HTS data, a number of CNV detection algorithms have been developed. Since HTS data contains various types of biases, these algorithms usually have a bias correction step. However, these bias correction methods are often large bin-based and the resolution of these algorithms is heavily restricted by the bin size. Here, we developed an algorithm that can normalize HTS data at a nucleotide level as well as a CNV detection algorithm based on the normalized HTS data that can detect CNVs with base-pair level resolution. Simulation and real data analysis shows that this algorithm can effectively remove the biases and accurately call CNVs.