Once a file is compressed to a JPEG format, JPEG noise is added to it (or substracted from the point of view of details). If you reopen this JPEG file and save it with a higher quality setting as JPEG or to any other format, the noise will be saved with it. So converting a JPEG file to a PNG won't remove the JPEG artefacts.

JPEG noise looks like little squares that typically appear around sharp edges and loss of details in textured areas. They are due to the way JPEG compresses image data by truncating the Discrete Cosine Transform (DCT) coefficients. This noise is often called JPEG artefacts since it doesn't have typical noise properties such as randomness. Here below are a few examples of a very close view of image portions with and without JPEG noise. The image portions are taken from the lighthouse example. The image was compressed with a quality of 50. See image on the right. See the zoomed portions below.

JPEG
No noise

Looking at these close up pictures, it's clear that while the original image looked crisp, the edge have been corrupted with noise and more importantly some details have been lost, like for the wall area where the texture is totally blurred.

Imagiris is based on a new algorithm developed internally that completely breaks apart from existing techniques. The basic idea is very simple and isn't new: simply smoothen areas around edges without destroying the edges. That was the obvious part. But Imagiris goes much further than that by detecting the compression settings and automatically adjusting the smoothing effect to both the local structure of the image and the compression settings. This way, it's totally automated and the you doesn't have to adjust many esotheric knobs. Not to mention that by doing it algorithmically, Imagiris can determine the optimal settings instantaneously.

One direct benefit of automatic compression detection is that imagiris is going to preserve fine details (like for example skin texture) if they were preserved by the JPEG compression.

Note that the detection is not limited to JPEG images. If you compress a JPEG image with a quality of 50, then save a copy of it as PNG file (which obviously will still contain the JPEG noise and artefacts), Imagiris is still able to retrieve the original compression settings of the initial JPEG file.

The problem with details that don't present too much regularity is that once they're lost, they're lost for good. Well that's partially true it depends on their self-similarity and how much they're lost, and thus on the compression settings. New advances in fractal analysis allow to restore a lot more details than what was previously possible. This part is still in active development and all enhancements have not yet been made available through on-line access. But expect them soon.