Bad Pixel Replacement

[EDIT: 20110602 14:20 CDT - reformat - WDR]

I am trying to replace bad pixels in an hyperspectral image. In the code so far i have read the bad pixel file. I then created a matrix in [j,k] or sample and band space of 1's where pixels are bad and 0's where good. Created a linear index of of all the bad pixels (1's). Then I assigned j,k subscripts to the index. below is the code. Everything works so far. I am wondering how I can apply my bad pixels, subscripted by j,k to a 3-D hyperspectral cube in i,j,k space. Also, for each bad pixel need to take the weighted average of the adjacent good pixels. If two bad pixels in a row need to take 2/3 of the value of the closest good pixel, 1/3 of the value of the farther away good pixel. I need to be able to extrapolate on this process for any number of sequential bad pixels. Any ideas, specific or general, would be greatly appreciated.

badPixelMapName = ['bad_kh_082910' ];
filename = [badPixelMapName '.flt'];
hdrname = [badPixelMapName '.hdr'];

header = readENVIHeaderStruct(hdrname);
nb = header.lines;                            %in the header, lines are actually number of bands
ns = header.samples;
nl = header.bands;

% create 2D matrix (samples and bands) of bad pixels, with 1's representing bad pixels
fid=fopen([filename]);
fpixel=fread(fid,Inf,'single');
fclose(fid);
badpix = zeros(nb,ns);        
cc=0;
for j=1:ns;
    for i=1:nb;
        cc=cc+1;
        badpix(i,j)= fpixel(cc);
    end
end

% create linear index of bad pixels
ind = find(badpix);

% assign column(J) & band(K) subscripts to each of the bad pixels 
s = [102,640];
[J, K] = ind2sub(s, ind)

Answer 1

[Sean de Wolski]

What about using John's inpaint_nans?

http://www.mathworks.com/matlabcentral/fileexchange/4551-inpaintnans

http://www.mathworks.com/matlabcentral/fileexchange/21214-inpainting-nan-elements-in-3-d