This is the comparison sample that was cheap and took more photos in the previous single evaluation. Because the author naively thought that it was difficult to open the gap between mobile phones in the daytime, almost all of them were night shots and samples under weak light.
If you have or are going to buy one of Xiaomi 4/Moto X/Samsung Galaxy S5, I hope this article can help you get a general understanding of the night shooting performance of the machine.
Parameters and description
motoX： Sony IMX135 sensor, 13 million pixels, 1/3.06 inch, 1.12 μ m long single pixel, F2.25 aperture. The testing machine is the 32G version of the Bank of China, and the system is Android 5.0.2;
Xiaomi4: Sony IMX214 sensor, 13 million pixels, 1/3.06 inch, single pixel length 1.12 μ m, F1.8 aperture, 6P lens. The test machine is mobile 4G version, and the system is MIUI 6;
Samsung S5: 16 million pixels Samsung's own ISOCELL, 1/2.6 inch, with a single pixel length of 1.12 μ m. The test machine is Snapdragon 801 national travel version, which has been painted with Android 5.0. (Officially, ISOCELL can reduce the interference between pixels, improve the light input of single pixel, and improve the weak light performance)
This part of the sample was left in the MotoX evaluation before, because we knew that the night shooting ability of MotoX was not excellent, so we found Xiaomi 4 and Samsung Galaxy S5 as the comparison objects.
The left is the original image, and the right is HDR mode
First, I want to explain the night view mode of MotoX. Because it is an almost pure native system, the camera does not provide night view mode. The author uses HDR mode as a substitute. After HDR is turned on, the brightness of the image has been improved by a gear, and the brightness is higher than that seen by the naked eye. Although the image still has a lot of noise, the color of the night scene sample will appear more pleasing.
Original drawing
The central part is magnified by 100%
Original drawing
Central magnification 100%
The center of the picture has been enlarged by 100%
In addition to improving the brightness of the image, HDR also smeared the super serious noise to a tolerable level, but it failed to improve the discrimination of the machine night scene, and even caused some details to be lost due to smearing noise (words on the ice-cream cone at the bottom left of the above figure). However, since its advantages far outweigh its disadvantages, it finally decided to let Moto X participate in the night scene comparison in HDR mode.
Note: MotoX is on the left, Xiaomi 4 is on the middle, Samsung S5 is on the right, and the three machines are in night view mode.
In this group of samples, Xiaomi 4 in the middle forgot to open the night view mode because the author is cheap
Ultra weak outdoor lighting, the ISO of MotoX and Samsung S5 has soared to 1000, and that of Xiaomi 4 has soared to 1300; The shutter speed is very close, but the 1/17s of Samsung S5 is faster and the film rate is higher.
In terms of picture brightness, motoX is a bit higher, followed by Samsung S5. Here Xiaomi 4 has no night view mode, and the brightness is the lowest. In terms of noise, motoX successfully wipes out the ground noise, but is powerless to the sky; On the contrary, Samsung S5 has solved the noise in the night sky, but the ground and scenery noise are relatively obvious; Surprisingly, Xiaomi 4, the highest in ISO, has the best noise control.
In terms of discrimination, the Samsung S5 is the only machine that can clearly see the bamboo leaf outline beside the vehicle in the center of the picture, but this is more due to the accuracy of focusing and algorithm, and it is the worst one in terms of floor details.
Although motoX has the most noise, it retains the most details and has the best analytical power (the steps of the escalator and the artificial waterfall on the left have the most details); After Xiaomi 4 turned on night view mode, the brightness of the picture exceeded Samsung S5, but was still lower than Samsung S5, and the picture seemed to be coated with fog, but the noise of the picture was successfully removed; The Samsung S5 package tail, even after sharpening, has almost lost the details of the artificial waterfall and ground on the left side.
Take the road sign as the focus, and zoom in to the same size. The photometry of the three machines are very close, but at this time, the ISO of the motoX is 1600, the Samsung S5 is 1250, and the Xiaomi 4 has soared to 2654.
MotoX noise is exaggerated, but can barely recognize the words on the road sign; With IMX214 and ferocious daubing algorithm, Xiaomi 4 has the cleanest picture, but it makes the samples seem out of focus (all five samples taken at the same time are the same). Like the sharply sharpened Samsung S5, it is difficult to recognize the road sign font.
Original drawing
Central magnification 100%
The true ultra weak light, the ISO of MotoX is 1600, the ISO of Xiaomi 4 is 3265, and the ISO of Samsung S5 is 2000. It can also be seen here that the extreme safety shutter of the three machines is 1/15s for motoX, 1/14s for Xiaomi 4, and 1/17s for Samsung S5, which also verifies the actual performance of Samsung S5 with the highest film yield.
In this extreme environment, the three machines are already using the highest ISO, so there is no gap in the picture details. MotoX is still heavily noisy, S5 is full of graininess after noise is smeared and sharpened, and Xiaomi is heavy handed.
In this complex light environment, the white balance of the three machines is quite different. MotoX is the closest to reality, but the green is surprisingly bright (grass in the lower left corner); Xiaomi 4 is blue, while Samsung S5 is yellow.
The sidewalk is focusing in the distance, and the sample is magnified by 100%
The white balance of motoX is warm and yellow, millet 4 is slightly cold and blue, and S5 is largely green. On the first look, S5 is the best detail, but it is actually only the contribution of sharpening (the details of leaves in the lower left corner and the upper right corner have not been sacrificed). MotoX without sharpening treatment and with night scene pattern painting retains more details; Xiaomi 4 still seems out of focus.
Enlarge the center of the picture to the same size
Here, Samsung's furious algorithm has been rewarded. The ice-cream cone in the lower left corner and the reflector in the middle have the best detail performance. MotoX is still lost to noise, and Xiaomi 4 is still lost to "algorithm paste".
Original drawing
Central amplification
Zoom in at the lower right corner
True extreme weak light, the three machines all use the highest ISO, and maintain their own safety shutter. In terms of the overall brightness of the picture, motoX is still the highest (the brightness is much higher than that seen by the naked eye), followed by Xiaomi 4 and S5 (the closest to that seen by the naked eye). MotoX sacrifices noise control, but it has the best image appearance and retains the most details.
Summary
The night imaging characteristics of the three machines are very distinct:
MotoX sacrifices noise control for brightness. It will be brighter than the naked eye in any shot. It can retain most of the dark and picture details, but the bright details are doomed to be sacrificed;
Xiaomi 4 originally had the problem of algorithmic painting during the day, but after the night view mode was turned on, it was even more exaggerated. Although the picture is the cleanest and the details are kept the least, it even gives people the illusion that they are not focused;
Samsung S5's ISOCELL is severely ineffective at night and lacks details, but it has Samsung algorithm bonus. After heavy smearing and sharpening, it shows sharp details at first glance, but in fact, details are lost seriously.
If the author was asked to rank their night shooting ability, the result would be: motoX>Samsung S5>Xiaomi 4.
Although motoX noise is flying all over the sky, it is better than the high brightness of the picture. The first impression is good. After ignoring noise, the details of brain compensation are not bad, and the experience of fast imaging is good; The ISOCELL weak light of Samsung S5 is congenitally deficient, and there is no way to recover even if there is a frenzy algorithm. The details that cannot be captured by the sensor are impossible to be solved by the algorithm; The most regrettable thing is that Xiaomi 4, with the support of IMX214, has smeared out all the details that are hard to capture.
