If you're wondering if there are noise modes 1-6, there indeed are. I'll introduce them next time, although most of them don't come into play for a few posts, yet.
Sorry for taking so long to read this. It's very interesting.
Intuitively, would you say it feels wrong to you for the trees image to have lower ESC than the Bentley photo? I think I probably would say so.
I suspect the compression curve under noise mode 7 mainly just reflects shortcomings with the compression method. Did you try using a more lossy compression for it?
It's almost like we're trying to reverse engineer the compression processes I imagine underlie our intuitive experience of "complexity".
I think the measure seems to work very effectively/as expected for more simple images, such as flags and cartoons. I'm not entirely sure what to make of that fact.
Heh, no worries about reading, they aren't going anywhere. I'm up to part 6 and will try to get out part 7 this coming week.
Yes, on the presumption ESC measures complexity, I agree the low value with "Trees" is surprising. I *think* what it's telling us is that the image is already in the "random" zone. It's already "noisy". It has notably low compressibility.
FWIW, I tried that image, with mode 7 noise, as PNG and JPG files. Info in this Note:
I think the main problem with mode 7 is that it affects all pixels with the percentage of "noise" being how much the pixels get affected. It's interesting how all three have different values. Another interesting thing is that I can see the lossiness of the JPG images in the higher noise values. Definite difference from the BMP or PNG images.
I do think compression is at the center of this. To some extent the SS and EF values seem most representative of that. The ESC is strange, though, and I still haven't figured out what to make of it. I think it makes more sense in a given domain — for flags or for cartoons, as you say. It varies so much with file size that I think it's hard to use across a broad spectrum of texts or images.
If you're wondering if there are noise modes 1-6, there indeed are. I'll introduce them next time, although most of them don't come into play for a few posts, yet.
Sorry for taking so long to read this. It's very interesting.
Intuitively, would you say it feels wrong to you for the trees image to have lower ESC than the Bentley photo? I think I probably would say so.
I suspect the compression curve under noise mode 7 mainly just reflects shortcomings with the compression method. Did you try using a more lossy compression for it?
It's almost like we're trying to reverse engineer the compression processes I imagine underlie our intuitive experience of "complexity".
I think the measure seems to work very effectively/as expected for more simple images, such as flags and cartoons. I'm not entirely sure what to make of that fact.
Heh, no worries about reading, they aren't going anywhere. I'm up to part 6 and will try to get out part 7 this coming week.
Yes, on the presumption ESC measures complexity, I agree the low value with "Trees" is surprising. I *think* what it's telling us is that the image is already in the "random" zone. It's already "noisy". It has notably low compressibility.
FWIW, I tried that image, with mode 7 noise, as PNG and JPG files. Info in this Note:
https://substack.com/@wyrdsmythe/note/c-137151533
I think the main problem with mode 7 is that it affects all pixels with the percentage of "noise" being how much the pixels get affected. It's interesting how all three have different values. Another interesting thing is that I can see the lossiness of the JPG images in the higher noise values. Definite difference from the BMP or PNG images.
I do think compression is at the center of this. To some extent the SS and EF values seem most representative of that. The ESC is strange, though, and I still haven't figured out what to make of it. I think it makes more sense in a given domain — for flags or for cartoons, as you say. It varies so much with file size that I think it's hard to use across a broad spectrum of texts or images.