Introduction

Demosaicing is a technique to reconstruct the low resolution image from image detector to the high resolution one. Typically, digital color cameras use a single image detector with a color filter array (CFA). In this design, every pixel on the detector collects the intensity of only one of the many color channels. To recover the full color image from the CFA-based detector, a method is required to estimate the missing color information. These methods are referred as demosaicing algorithms.

Figure 1: Demosaicing

Many demosaicing algorithms have been surveyed in previous work[1]. In this project we studied and implemented four prevailing demosaicing algorithms. They are:

Pixel-Value Data-Dependent Triangulation[2]
Two Line Interpolation[3]
Edge Directed Demosaicing[4]
High Quality Linear Interpolation[5]

We also implemented the bilinear interpolation, a straightforward demosaicing algorithm, to help evaluate the performance of the new algorithms. Most of these algorithms are specifically applied to Bayer CFA[6], the most commonly used CFA. In coming sections we will describe these algorithms, compare the results, and make a conclusion about these algorithms.

< HOME > < Algorithm Description >