PowerPoint File for Presentation:

Algorithm Implementations:

BilinearInterpolate.m:

function IMG_Bilinear = BilinearInterpolate( IMG );

Bilinear interpolation that avearage existing adjacent pixel to fill in the missing color information.

DDTriangulate.m

function IMG_DDT = DDTriangulate( IMG );

Data-Dependent Triangulation[1].

EdgDirInterpolate.m

function [IMG_ED, EdgDet_map] = EdgDirInterpolate( IMG )

Edge Directed Interpolation. Using the variance of the surrounding pixels to reconstruct the missing color information[2]. EdgDet_map is the debug information that shows the pixels classified as an edge.

HiQualLinearInterpolate.m

function IMG_HQL = HiQualLinearInterpolate( IMG );

High quality linear interpolation[3]. Involved some gradient idea to help refine the bilinear interpolation result.

TwoLineInterpolate.m

function IMG_TwoLine = TwoLineInterpolate( IMG )

TwoLineInterpolate uses only 1 line buffer to implement the interpolation. It uses the R/G and B/G ratio to help determine the local gradient and recover the missing color information[4].

Other Functions:

Demosaicing.m

A script running all algorithms. Set the algorithm variable to 1/0 and it will be excuted/disabled. Loaded test image can be changed by reassigning "filename", "filefmt" at line 19,20.

GrayImageRecover.m

Recover the gray image (MxNx3) from a CFA image (MxN). Used to reconstruct specific test patterns.

ImgComp.m

function [DIFF, ERROR_DIST_IMG] = ImgComp( Input, Ref, Method);

ImgComp compares 2 input images "Input", "Ref" according to "Method" and returns the result value and the error distribution map.

Interpolate2L.m, Interpolate2Lfactor.m

Two sub functions used by two-line interpolation.

RGB2CFA.m

function [Result_1D, Result_3D] = RGB2CFA(Img_RGB)

RGB2CFA downsamples the input image to a CFA image using RGGB format (The 1st pixel in the 1st line is R) Both 1-channel image and 3-channel image are available.

< Home >