Image demosaicing by an improved alternating projections algorithm based on dual-tree complex wavelet transform

Most digital cameras acquire image information as an incomplete down-sampled representation of the three basic color components, such that one color value is sampled per pixel location. Image demosaicing is the process of retrieving the missing two color components in each pixel. Demosaicing by alternating projections (AP) is one of the most successful demosaicing approaches. The contribution of this paper is developing enhancements to the existing AP algorithm by proposing the dual tree complex wavelet transform as a framework for the algorithm, strengthened with wavelet-based denoising to achieve better color channel reconstruction. The proposed enhancements result in an average 0.27 dB PSNR improvement as tested over the Kodak Photo CD set. Visual experimentations also verify this PSNR raise. The enhanced AP algorithm is also shown to perform comparably to the leading demosaicing techniques.