Skip to Main Content
Most cost-effective digital camera uses a single image sensor, applying alternating patterns of red, green, and blue color filters to each pixel location. Demosaicing algorithm reconstructs a full three-color representation of color images from this sensor data. This paper identifies three inherent problems often associated with directional interpolation approach to demosaicing algorithms: misguidance color artifacts, interpolation color artifacts, and aliasing. The level of misguidance color artifacts present in two images can be compared using metric neighborhood modeling. The proposed demosaicing algorithm estimates missing pixels by interpolating in the direction with fewer color artifacts. The aliasing problem is addressed by applying filterbank techniques to directional interpolation. The interpolation artifacts are reduced using a nonlinear iterative procedure. Experimental results using digital images confirm the effectiveness of this approach.