doubt on using blkproc functiom

hai i am working with matlab R2009a and i have to move a window of size[30,700] over the image and calculate the sum of pixels inside the window

the problem is my image is having NaN values and and i am using the blkproc function to make the window move over my image and i got a matrix of size 21X700 with NAN values is my answer correct please some one help me

and my coding is this

figure ,imshow(final2); [width, height] = size(final2); fun=@sum; B=blkproc(final2,[30 height],fun);

Answer 1

[Andrei Bobrov]

variant

B=blkproc(final2,[30 height],@(x)sum(x(~isnan(x))));

OR

B=blockproc(final2,[30 height],@(block_struct)sum(block_struct.data(~isnan(block_struct.data))));

add variant

B=blkproc(final2,[30 height],@(x)sum(x(~isnan(x)&~isinf(x))))

[ramya raj]

hai i have used the

B=blkproc(final2,[30 height],@(x)sum(x(~isnan(x))));

but i am getting the answer in a 21X1 matrix as all NAN,Inf,0

what is the problem i can't understand

[Andrei Bobrov]

Give part of your data

[ramya raj]

hai this is the data that i have

and the code i have used

clc;

clear all;

close all;

fundus=imread('D:\10mcs010\im0001.ppm');

grayImage=fundus(:,:,1);

[rows columns numberOfColorBands] = size(grayImage);

figure,imshow(grayImage);

title('Original Grayscale Image');

binaryImage = grayImage > 35;

A=[0 0 0;1 1 1;0 0 0];

op=imopen(binaryImage,A);

clos=imopen(op,A);

eror=imopen(clos,A);

maskedRgbImage = bsxfun(@times, fundus, cast(eror,class(fundus)));

g=maskedRgbImage(:,:,2);

imv=conv2(double(g),[1 0 -1],'same');

figure,image((imv));

title ('vertical image');

imh=conv2(double(g),[1;0;-1],'same');

figure,image((imh));

title ('horizontal image');

ev=abs(imv);

eh=abs(imh);

edgediff=imsubtract(ev,eh);

edgediff=double(edgediff);

erore=double(eror);

g=double(g);

final2=edgediff ./erore;

figure ,imshow(final2);

[width, height] = size(final2);

B=blkproc(final2,[30 height],@(x)sum(x(~isnan(x)&~isinf(x))));

and this is the answer i am getting i don't know where is the error

-44907

-13744

-9499

-1979

822

1502

-3559

12038

6991

14429

9999

8073

9060

4894

-3508

-8374

-51

-3687

-1323

-38133

0

Answer 2

[Bjorn Gustavsson]

Replace sum with nansum in your call to blkproc. If you dont have the statistics toolbox you can look for nansum in the file exchange - I've found at least one statistics-related toolbox there that includes a number of nan-safe functions.

HTH