Image processing before the interview, the interviewer asked such a question:
Two of the same type camera respectively in different scene took a photograph, comparing two pictures who more clearly?
Which the human eye can separate pieces of fuzzy, but how to design algorithms identify? Study database will not be used ()

Definition of the general algorithm or image quality assessment are for the same scene, different scenarios for gradient, variance, there is no comparability, such as information entropy so I don't really have ideas, cool the interview, the interviewer didn't and I said, we help people have any idea about this?

CodePudding user response:

Read the image transformation to the frequency domain, frequency component of the image, clear image of the high frequency component is more, the fuzzy images of the low frequency component is more,
After the image into grayscale, with Sobel or Laplacian processing, to calculate the average grey value of image, the greater the value, on behalf of clear image, on the other hand is fuzzy,
Or gray image, the image gray variance, the larger the variance, the higher the said resolution

CodePudding user response:

Different scenarios can not compare, it is difficult to the same scene, there are some subjective content, such as: face is the background of the shutter, if the focal length of the latter, so Sobel return energy or similar algorithm is higher than the former, but it is burnt to even run, in the photography
In addition to improve the sensitivity, turn off the noise reduction function at the same time, the value will be bigger, in fact it is caused by noise,

CodePudding user response:

Different scene image sharpness can compare? If there is no way