MATLAB爱好者论坛-LabFans.com - 查看单个帖子 - Speeding up Matlab Gaussian Filter implementation

poster · 2019-11-25, 05:20

I am un familiar with matlab. I have calculated a kernel of size NxN refered to as mask in the code below. I then pad the image, before applying the convolution operation seen below.

for i=1:size(temp_image,1)  %temp size of output image

   for j=1:size(output_image,2) %size of our output image

     temp = double(padded_input(i:n-1+i,j:n-1+j)).*mask; %padded_input is our padded image

     temp_image(i,j) = sum (temp, 'all');

   end

end

This convolution calculation adds significant time to the operation, notably 3 seconds for my current test case. How can I reduce this convolution operations processing time? I would hypothesis there is a more efficient way to do this calculation.

Thank you

2019-11-25, 05:20	#1
poster 高级会员注册日期: 2019-11-21 帖子: 3,006 声望力: 66	Speeding up Matlab Gaussian Filter implementation I am un familiar with matlab. I have calculated a kernel of size NxN refered to as mask in the code below. I then pad the image, before applying the convolution operation seen below. `for i=1:size(temp_image,1) %temp size of output image for j=1:size(output_image,2) %size of our output image temp = double(padded_input(i:n-1+i,j:n-1+j)).*mask; %padded_input is our padded image temp_image(i,j) = sum (temp, 'all'); end end` This convolution calculation adds significant time to the operation, notably 3 seconds for my current test case. How can I reduce this convolution operations processing time? I would hypothesis there is a more efficient way to do this calculation. Thank you More answer...