No, it does not. You need some kind of converter.
However, why do you want to do that? Converting to PNG does not magically recover the quality lost to JPG.
Of course not, but it stops further loss of quality.
"further loss of quality" for a JPG only makes sense if you edit the photo in an editing suite, make changes and save it again, and choose JPG deliberately. Storing, copying, and viewing JPG files doesn't change them at all. Converting to PNG just gives you a PNG that looks exactly like the JPG. So if you've got jpg photos then you can just leave them as JPGs and not worry about any loss of quality: the loss of quality was a one-time thing when the file was first created.
The only time they can get further degraded is if you open them in a photo-editing program, and save them again as a lossy format, especially if the format isn't exactly the same as the original JPG, or if you make changes such as applying filters to the whole image, or cropping it (since that usually throws the JPG blocks out of alignment). The rule of thumb here is to just leave your image files alone as whatever format they happen to be in, but remember to choose PNG for saving, when you edit pictures in a paint program. It makes literally zero difference if you convert them to PNG before editing.
~~~
Some of the new AI-based filters might help out in the near future. The premise is that they train the AI on millions of "sample" pairs of high-quality / low-quality versions of the same images, teaching it to repair the damage on the low-quality images so that they look as much like the high-quality images as possible.
You can then use the resultant neural network to clean up other images. However, the closer the training set resembles the data you wish to clean, the better. e.g. if you had low-quality JPGs of anime images, and wanted to upscale them while removing JPG compression artifacts, you'd want to get a huge number of high-quality anime images, downscale them, then convert to JPG, with as close to the same settings as possible as the images you wish to clean up. Then teach an NN to reverse that specific degradation for that specific class of images, and you'd have a custom filter which was really, really good at repairing images exactly like those ones.
EDIT the concept here is that if you teach an NN to upscale images of e.g. ... WWII planes, then it becomes really good at working out what WWII planes are
meant to look like, so when you give it a new image of an unknown WWII plane, it fills in missing details intelligently, rather than just using a generic upscale filter. Sure, it will still upscale other types of things and might do passably ok, but it's going to be really good at the class of things it was trained on. In the future there might even be intelligent online filters that work out
what something is, e.g. object and edge-detection, then apply specifically trained filters to subregions of the photo which hold those things. e.g. if there's a blurry plane and a blurry tree, it could apply different upscale filters to both regions of the photo, and do better than applying only one filter to the whole photograph.
Author
Topic: The Generic Computer Advice Thread (Read 573439 times)