When checking the duplicate, you can identify the table contents that are not in the form of pictures. If you cut the table into a picture, there is no guarantee that 100% cannot be recognized. At present, HowNet can recognize pictures, so cutting tables into pictures only increases the difficulty of recognition and cannot be completely avoided.

Published papers can be in the form of pictures, but this will lead to a reduction in the number of words in the paper. If you use a lot of screenshots, it will definitely affect the number of words in the whole paper and the results of duplicate checking. As for whether it is allowed or not, I suggest you consult the journals to be published.