1. 31
  1.  

  2. 16

    This isn’t merely blurring. It searches for an image without blocking artifacts that would compress to the exact same data as a given JPEG.

    1. 4

      I used this some years ago.

      There are several others like it and IIRC one of the others was drastically faster, but when I needed something like this I picked jpeg2png because its choices are governed by a reasonable mathematical model. Noone has complained, which might perhaps mean that jpeg2png is great. YMMV.

      1. 3

        Interesting approach. I played around a bit with it, and unfortunately it tends to destroy high frequency content. This is somewhat expected, and there’s really only so far you can go with baseline JPEG.

        1. 3

          Isn’t high frenquency content destroyed by the encoding. So this thing doesnt destroy the high-freq, it’s just that the data isn’t there. And any high contrastyness in the comparative jpeg isn’t data, but just an artifact. So it’s not right in this case to blame jpeg2png.

          1. 1

            Yes and no. It’s behaving similar to encoders that try to minimize PSNR. It does not take any structure in the image into account. This is most visible in images with a lot of salt & pepper in them. Another way to state this is that it would be better if jpeg2png had a psychovisual model of some form.

        2. 5

          This is very cool. It’s unfortunate that the very first example is the Lena image, though. :-(

          1. 7

            I wish people would stop using that image as an example. Not because of its content—though that part isn’t ideal either—but because the source file is so old and mediocre that it isn’t remotely representative of the images we routinely handle today. Worst of all the extreme colour cast makes it useless for judging skin tones.

            1. 4

              The author doesn’t seem to be from the US, and the last commit appears to be from 2016.

              1. 3
                1. 3

                  I find very unfortunate that there are people who find this unfortunate. You are literally spreading cancer culture.

                  1. 5

                    I agree, the whole “Losing Lena” movement was a total beat-up of what is really just a small number of people using this image essentially as a in-group meme. The cropped version (the only one I’ve ever seen used in the past 20+ years) is ridiculously tame compared to images that continuously bombard us in contemporary media, especially marketing targeted at women.

                    That’s not to say we should continue using the image. We shouldn’t. We should stop using it because it’s grossly unrepresentative of photographic depictions of human skin. We should stop using it because the image is horrifically poor in quality. We should also stop using it because there’s no reason to use something with even a jot of sexual content.

                    1. 4

                      No one is being cancelled, though?

                      1. 3

                        I mean… I’m pretty skeptical of a lot of this stuff, but afaik Lena has personally asked people to stop doing it. It’s not some “hypothetically she might be upset”, it’s “the subject of this picture has requested that you find an alternative”.

                        1. 3

                          Do you have a reference on Lena asking people to stop using it? I couldn’t find anything in that Wikipedia article, and it would change the situation considerably.

                          I mean, regardless of whether she actually said anything or not, I think I would be on the side of “maybe we shouldn’t be using that image everywhere”. A lot of my work involves working with raw pixel data, and I find the tool at https://rawpixels.net super useful. However, it uses the Lena image as a default/placeholder image, and looking at very obviously sensual pictures of women, who show no signs of having clothes on, on a 28” display at work, in an open office space, is pretty awkward.

                          1. 10

                            I’m Lena. I retired from modeling a long time ago. It’s time I retired from tech too.

                            https://vimeo.com/372265771

                            1. 0

                              Do you have a reference on Lena asking people to stop using it? I couldn’t find anything in that Wikipedia article, and it would change the situation considerably.

                              Hrm. I recalled it was when she went to the “Conference of the Society for Imaging Science in Technology”. I’ve gone and dug up what she actually said, and it would take a real effort to twist it to the interpretation I’d heard. Sounded to me more like she’s bemused by the popularity of the image than upset.

                              Deliberately not reposting what she said here as I’m disinterested in a long thread of people dissecting it.

                              1. 4

                                She does ask for folks to stop using her photo for this purpose in the video at https://www.losinglena.com/.

                                1. 0

                                  Ah. I didn’t watch it because I dislike getting info via video and they’ve declined to offer any other format.

                                  Can you suggest a timestamp to make checking easy?

                      2. 1

                        PNG output is somewhat dangerous (lossy passing as lossless).

                        This risk can (and imho should) be managed aggressively in many ways, such as tags in the output file headers and file naming.