Quote from: Ricci Rox on April 27, 2018, 13:28:54
Hey there, S. YU. Article author here. Honestly, I've seen the P20 Pro do some pretty incredible things in low-light conditions. I'm not about to claim that it's a poor camera device. However, I've also seen some less than praise-worthy images taken by the device.
It's a bit of one, and some more from another, IMO. It's capable of great photos but does flounder in certain departments.
Hello Ricci,
It not really the point whether they directly paid DxO for the review or if it's just that DxO routinely flatters the newest model of those who buy their equipment. They may be professional, but they're not credible, it's clear from both the 808 review and the P20Pro review, also from some of their paradoxical results from their lens and camera body reviews, and I believe there certainly will be a lot more if somebody takes the time to actually look at their images.
Let me summarise the P20Pro's performance properly, and with proof I can't provide the proof due to the system blocking my links, I'll try to circumvent it:
1. Due to oversharpening and smearing NR, it falls short consistently (no exceptions compared to S9+ and Pixel2XL, probably even U11+ and Nokia 808, AFAIK) on detail and texture in daylight in both 40MP and 10MP mode as long as zoom is sub-3x.
cdn(dot)gsmarena(dot)com(slash)imgroot/reviews/18/huawei-p20-pro-vs-samsung-galaxy-s9-plus/camera/gsmarena_205.jpg
(same as above)/gsmarena_305.jpg
Note the snail, where the P20Pro loses a large amount of detail and renders the snail almost wax-like. In the distance, S9+ also retains much more detail of the castle.
cdn(dot)gsmarena(dot)com(slash)imgroot/reviews/18/huawei-p20-pro-vs-samsung-galaxy-s9-plus/zoom/gsmarena_166.jpg
(same as above)/gsmarena_366.jpg
Note how the branches are black and yellow and totally devoid of any detail in the P20Pro sample while S9+ at least still renders the branch properly as brown.
Same case in Anandtech's samples and Allaboutsymbian's.
2. Its 3x zoom gains an inconsistent advantage depending on light availability and possibly other factors. I've seen one review where it seems to retain even more detail than what 3x@8MP would suggest(a Chinese review), one review where it performs generally as expected(GSMArena's), and one review where there's hardly any advantage(I can't recall, it was at least 2 weeks ago), but I don't have the time to go back and look of them now. Anyway that's compared to modern flagships, compared to the 808 it seems yet inferior in good light.
allaboutwindowsphone(dot)com(slash)features/item/22915_Zoom_PureView_and_more_Lumia_1.php
www(dot)allaboutsymbian(dot)com(slash)features/item/22923_By_popular_demand_Nokia_808_Pu.php
3. As available light diminishes it gains in auto mode first and foremost DR but then gradually sharpness and noise advantage, it takes really dark scenes for a clear advantage, otherwise it mostly retains highlight DR better but renders a faker and less detailed image.
images(dot)anandtech(dot)com(slash)galleries/6261/p20pro_img_20180327_173028.jpg
(same as above)/s9plus_20180327_171413.jpg
In this dim indoor shot the S9+ retains notably better texture and detail of objects like chairs, the blackboard, the various plants and flowers, with minimal artifacts, while the P20Pro only has more highlight DR, with artifacts like strange black patches on the blackboard.
4. The 40MP mode and arguably the 3x zoom are worthless in low light. Low light mode(multi-frame exposure) should be avoided unless one just can't do a semi-auto image alignment and exposure merge in PS and really needs something in extremely dark conditions, otherwise it smears like there's no tomorrow and tends to overexpose the sky and introduces artifacts.
www(dot)phonearena(dot)com(slash)news/huawei-p20-pro-apple-iphone-x-google-pixel-2-xl-samsung-galaxy-s9-plus-low-light-night-camera-comparison_id104179
The 40MP shot in this article is indeed funny.
images(dot)anandtech(dot)com(slash)galleries/6261/p20pro_img_20180327_173048.jpg
Typical smearing watercolor.
cdn(dot)gsmarena(dot)com(slash)imgroot/reviews/18/huawei-p20-pro-vs-samsung-galaxy-s9-plus/camera/gsmarena_114.jpg
Again severe smearing and also some ghosting on the left of the image.
cdn(dot)mos(dot)cms(dot)futurecdn(dot)net(slash)EorYdWTrqXHZEJFxbM4v75-650-80.jpg
Overcooked HDR and unrealistic sky exposure evident even on such a small sample.
There's one shot in SF on one of the city's famous slopes about the sky exposure issue but I can't quite find it at the moment.
5. The camera generates the most "diverse" set of artifacts anywhere. From sharpening halo, smearing, random dark blotchs, to CA, interpolation artifacts, and possibly astigmatism.
Sharpening halo and smearing are pretty much everywhere, but auto mode smearing and night mode smearing are still on different levels. The dark blotchs appear in the shadows of low light shots. CA is apparent, consistently in 40MP output, just check out the high contrast edges of any 40MP sample. Interpolation artifacts are rampant in 40MP(again shown by GSMArena, their choice of testing ground is very indicative of a variety of performance measures) and 10MP is, surprisingly, not entirely immune. See the snail shot again, the building in the background has holes on outer panels which are not correctly rendered on the far side (near the castle in the image), there's a 40MP photo of the same panels which show much worse artifacts but I can't find it right now. And even more sinister is that sometimes it "invents" detail due to the interpolation without you knowing. An earlier sample also with that snail showed that the pattern of chipped paint taken by the P20Pro was completely wrong, there was a pattern not immediately strange to the eye but when compared to proper bayer results reveals that it's fake and totally made up.
www(dot)gsmarena(dot)com(slash)piccmp.php3?idType=1
This tool suggests that the edges and corners of the P20Pro samples (both 40MP and 10MP, but the lower res sample is more noticeable due to less overall softness) have lower sharpness than the center, also the softness in the corners exhibit a slight "double image" most noticeable on the crosses on the background of the scene. This behavior suggests astigmatism, very rare on tiny phone camera lenses.
6. The "master AI" would be more appropriately named "amateur AI", as no professional would add such cheap amateurish filters to their photos. They claim the pseudo long exposure involves the "AI", but it doesn't, at least it doesn't need to. Auto alignment was already quite mature in PS without need of any machine learning, and many ISPs in compact cameras could also complete multiframe NR in camera, just without the HDR combined into one single process, probably because it's prone to generating intense fakeness without user control, but considering how fake that monument looked in cdn(dot)mos(dot)cms(dot)futurecdn(dot)net(slash)EorYdWTrqXHZEJFxbM4v75-650-80.jpg , Huawei obviously doesn't mind.