Performing sliding window for feature extraction on a test image is very slow. What is the reason

Question

Beulah A 2016-3-2

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/270951-performing-sliding-window-for-feature-extraction-on-a-test-image-is-very-slow-what-is-the-reason

评论： Beulah A 2016-3-2

I am new to Matlab. Please help me to clear the issue. I need to perform feature extraction on a test image by sliding the window on image of size 512x512. The window size is 70x30. First I am cropping the image and extracting the feature, the storing in a CSV file for further manipulation. Performing the said operation is very slow. Matlab shows the status as busy for long time. I could not understand why the process is very slow. How to make it fast. the code is:

img=imread('1.tif');
k=1;
for i=1:512-30
    for j=1:512-70
        crop_image = imcrop(img,[i,j,30-1,70-1]);
        toCsv(k,:) = HOG(crop_image);
        k=k+1;
    end
end

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2016-3-2

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/270951-performing-sliding-window-for-feature-extraction-on-a-test-image-is-very-slow-what-is-the-reason#answer_211948

在 MATLAB Online 中打开

imcrop is going to be less efficient than just indexing the image.

crop_image = img(j:j+70-1, i:i+30-1, :)

Could you confirm that you want to move the window by only 1 row (or column) at a time, not by full blocks? I suspect you end up re-calculating a fair bit, but I am not sure how HOG is calculated.

Is there a reason you are not using http://www.mathworks.com/help/vision/ref/extracthogfeatures.html ?