Are there algorithms or ready-made solutions for determining image quality?

Question

linux

I'm looking for an algorithm or script/console program to determine the quality of a jpeg image, no special accuracy is required.
I receive product data from an aggregator, often the photos it provides are of low quality ( example ), I could download photos directly from the manufacturer's website (aggregator provides a link) and replace bad photos.

Answer 1

comparefrom the imagemagick package, see for example .

Answer 2

There is no need for neural networks.
After all, the picture can initially be blurry, cloudy, or with a small number of details (for example, some simple object, such as a cube, on a plain background). And conventional algorithms will not be able to find out how much compression of the source material (if it exists at all). After all, you can open a picture in any viewer, enlarge it, take a screenshot, and you will get a completely high-quality image, in terms of jpg (and any other similar format).
However, already now there are ready-made networks that improve the clarity of the picture, and not only photos, but also videos, focusing on a large array of images from which they were trained, how to recognize where it is heavily compressed and where it is not, and what the original image can be. Modify it a little, and you can make a slightly simple solution - just determine how much the image is compressed (without reference to what format and what is written there, purely by the content of the image, as the human brain does).
Perhaps there are already ready-made solutions, I did not even google on this topic. The direction, I think, is clear.