I have the paper titled "CG-DIQA: No-reference Document Image Quality Assessment Based on Character Gradient". It is all about assessing the quality of the document image based on its blurriness. To compute the overall quality score s of a document image, below are the summary of the steps done:

  1. An eligible image character patch c is convolve with a horizontal filter fx and a vertical filter fy, which outputs an array of gradient magnitudes. The gradient magnitude of c at position (i,j) is denoted by mc(i, j).

enter image description here

  1. Get the average gradient magnitude of a specific document image denoted by ma, where N is the total number of pixels present in all character patches of a specific document.

enter image description here

  1. Lastly, to the compute the overall quality score s of a document image, standard deviation of the character gradients will be determined.

enter image description here

Now my questions are :

  1. What does it mean when standard deviation is small and big? Does it mean the document image is blurry when the standard deviation is small or vice versa?

  2. How can I normalize the standard deviation outputs , to give it a value of 1 when it is not blurry or less than 1 when its blurry.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.