How to extract sliding block features of an image by using blockproc? Please analyze my code

Hello dear all I am trying to extract sliding block features of an images by using blockproc. Firtsly i have tried nlfilter, but it is not useful sliding blocks by jumping 20 pixels right and down as Sean de Wolski and Image Analyst mentioned.

I have created a function to extract block properties. It worked with nlfilter but matlab gave some error for blockproc such below. Please tell me where im doing wrong and give an advice please. Thank you very much for your contribution...

Error message is: Function BLOCKPROC encountered an error while evaluating the user supplied function handle, FUN. The cause of the error was: Error using imhist Expected input number 1, I or X, to be one of these types: double, uint8, int8, logical, uint16, int16, single, uint32, int32 Instead its type was struct. Error in imhist>parse_inputs (line 277)

My code is

% blockproc Block Processing  
close all; clear all; clc;

global num;
global cnt;

cnt = 1;
num = 0;

I = imread('o1.jpg');
I = rgb2gray(I);

fun = @fn_blk;
disp('Performing blockproc Operation')
tic
B = blockproc(I,[210 210],fun,'BorderSize',[20 20], ...
   'PadPartialBlocks',true);
toc

imshow(I), title('Original')
figure
imshow(B, []), title('blockproc')

----------------------------------------------------------------

function fn_blk(x)

global cnt;
global inputs;

      count = imhist(x);
      % normalize histogram values
      count = count./numel(x);
      L = length(x);

      % calculate mean directly
      mu = centralmoments(count,3);

      inputs(1,cnt)= mean(x(:));                        %mean: average intensity
      inputs(2,cnt)= mu(2)^0.5;                         %standard deviation: average contrast
      varnorm = mu(2)/(L-1)^2;                          %smoothness: ranges from 0 (constant intensity) to 1
      inputs(3,cnt)= 1-1/(1+varnorm); 
      inputs(4,cnt)= mu(3)/(L-1)^2;                     %skewness of histogram: 0 for symmetric histograms
      inputs(5,cnt)= sum(count.^2);                     %uniformity: high value if all gray levels are equal
      inputs(6,cnt)= -sum(count.*(log2(count+eps)));    %entropy: randomness
      cnt = cnt + 1;

end