howto divide an image into K*K windows for LBP

Question

Newman 2016-7-11

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/294845-howto-divide-an-image-into-k-k-windows-for-lbp

编辑： Newman 2016-7-11

hello I have an image which is 112 x 92. I want to divide the image into 7x7 blocks and apply generate LBP pattern for each block that is for ieach of the 49 blocks.But the issue is 112/7=16 pixels as rows but 92/7=13.14 pixels as columns .SO I want to ask should I round it up to 13 pixels for columns ?Will it cause problems as generating a good feature vector for the face?

another question -I know how to create a global feature vector for the face, i need to concatenate all the blocks LBP patterns .So what should be the direction to do it for eg look at the image below:

Is this the correct direction or there is no direction we can concatenate all the blocks in any way we want ?

P.S- For my code/approach see below

image =112x92 unit8;

extract the size of image

     [rows col]=size(image);
     a=rows/7;
     b=round(col/7);
     dfrows=repelem(a,7);%repeating each element 7 times
     dfcol=repelem(col,7);
     dfcol(numel(dfcol))=dfcol(numel(dfcol))+1; %adding the last element with one as to equal 92 in the columns
     ced=mat2cell(image,dfrows,dfcol);

where ced is

    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]
    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]
    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]
    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]
    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]
   [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]
   [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x13 uint8]    [16x14 uint8]