jasle.blogg.se

Look at the eyes, for instance, or the feathers on the backs. There are limits to how well the upscaling can be done for complex subjects like birds, mud or water. The 4x upscale is too smudged and lacking in any detail. I actually like how Luminar 4 renders the mud a lot better. With abstract, chaotic patterns - for instance the mud - this algorithm falls apart and introduces weird patterns that look like piles of dead worms. Instead, it seemed to look for tiny details like the 'flow' of feathers on the tree pies' heads, and upscaled those patterns intelligently, rather than simply magnifying them as Luminar seems to do. I do not believe that the software recognized these as birds. The 2x upscale has significantly more detail than the original or upscale from Luminar 4. Viewed at 100%, things shift in favor of Gigapixel AI. I do not have a problem with the noise and actually believe it adds more to the photos. I actually like the results from Luminar 4 better because they preserve more detail. JPEG upscaled to 2500 px and exported from Luminar 4Ģx output from Gigapixel AI with the best settings (Art, 100 deblur, 0 noise reduction)Īt normal viewing size, all images look close to each other. The crops do not match exactly but it doesn't bother me.ġ:1 crop of the 16MP RAW file, processed in Luminar 4

I also used the Art setting because the default Standard setting introduced weird artefacts, as did the others.

The blur reduction improved sharpness and I wanted to keep noise to preserve detail as far as possible. A 4x upscale in Gigapixel AI with the same settings as 3.For me, that meant using the 'Art' algorithm with 100 blur reduction and 0 noise reduction

I downloaded Topaz Labs' Gigapixel AI and ran a quick test to see how it compares to my Luminar 4 software.įor this test, I took a 1:1 crop from a photo taken with my Canon FD 300mm f2.8L lens and my E-M5 (mark 1).