preface
When you buy a mobile phone, you will struggle with the ability of the mobile phone to shoot, which is closely related to the manufacturer's Lens (lens) design, related algorithms, Sensor (sensor) and ISP (image processor). Among them, Sensor (sensor) as an important basis for mobile phone imaging, its size, pixel design and design process will affect the final imaging to a certain extent.
Different Sensors, will the gap be very large? In order to verify this problem, the author found a mobile phone equipped with IMX700, IMX766, S5KHMX and GN1, in this article to do a test, these Sensors are very popular in 2021, mostly used as mobile phone main camera.
Below is a list of Sensors participating in the test, of which the largest size is the IMX700, up to 1/1.28 inch, the highest pixel is S5KHMX, up to 108 million pixels, all Sensors support pixel pooling to increase the size of a single pixel, improve the level of imaging.
(Information on the 4 different Sensors participating in this test)
This article has six links, which are to test the benchmark image quality of each Sensor, high-quality image quality, tolerance, night performance, white balance and color, and focus speed at different illuminations.
1. Benchmark image quality
The first is the benchmark image quality test, for the sake of fairness, the author will be each Sensor marked with the letters A-D, using the mobile phone's own professional mode shooting, in order to reduce the impact of AI algorithms on the imaging effect, with a heavy tripod and large gear gimbal to stabilize the equipment to ensure the stability of the picture.
When the picture exposure is normal, the lower the native ISO (sensitivity), the more delicate the picture quality. Since the sensor manufacturer did not announce the native ISO range of these Sensors, the author manually adjusted the ISO shooting supported by the phone (it is possible to use the extended ISO). The center image of the general lens is the best, but the different Lens design schemes of each manufacturer affect the extreme edge image quality, and the test picture will take the center and non-limit edge areas as a reference.
Test scenario: Resolution standard under indoor stable light source
(Test Scenario)
Central image quality is a very meaningful reference indicator, and the best image quality of Sensor can be reflected in the best imaging area of Lens. Since most mobile phones have a Sensor ratio of 4:3, the author aligns according to the digital id of the standard, which is also a test method to avoid light interference and visualize the quality of Sensor.
(Sample shooting)
Let's first take a look at the central quality performance of these Sensors at the lowest ISO, please observe the thin lines in the picture below and the sharpness and clarity of their intervals, as well as the overall contrast of the picture. Sensor A and D have the most solid picture quality and high contrast, and Sensor C's contrast and sharpness are good.
This is followed by a display of the lowest iso edge quality, and the enlarged area is located in the lower right corner of the picture, so that you can observe the sharpness of the numbers and the sharpness of the edges of the thin lines. Sensor D has the most solid picture quality, with higher contrast and sharpness, while Sensor A, B, and C have similar contrast ratios.
Second, high-sensitivity picture quality
At this point, you must have a certain understanding of the best picture quality of these 4 Sensors, but in the high-ISO scene when the light is not good, what is their picture quality? The highest range of ISO supported by these Sensors is 1600 to 12800, and to be fair, I use the 1600 ISO as a benchmark to test again.
It can be seen that when the ISO reaches 1600, the central quality of all Sensors has dropped to a certain extent, and the traces of smearing and sharpening are gradually revealed. Sensor D has the highest contrast and sharpness, but is darker.
The traces of texture smearing and sharpening in the marginal areas under the high iso are very obvious, but the lines of the individual Sensors are still recognizable.
Third, tolerance
Tolerance is also one of the important indicators to measure Sensor, the greater the tolerance of the Sensor, the more light and dark information is recorded, and the later adjustment is relatively more flexible. With the addition of HDR algorithms, does the law of "the bottom of the big family crush people" still apply?
Because most mobile phone users will not shoot RAW or high-pixel footage for post-processing
Most users often manually turn off HDR;
Most mobile retouching software does not support RAW format;
Most users' image processing is generally JPG pull exposure sleeve filter.
Therefore, the author used the JPG format of the default pixel size to shoot typical underexposed scenes, overexposed scenes and HDR scenes. Images are the minimum for each Sensor, and non-HDR scenes are shot in Pro mode.
Test scenario: Next to the window in the company aisle
(Xinghua students wrote "happy" on their faces)
The author caught a passing Xinghua classmate on the camera as a tool person to participate in the imaging test, and first took a sample of the normal exposure of the character. Before the adjustment, the sky is completely overexposed, and see if the details of a part of the sky can be pulled back through the corresponding post-production means.
I imported the JPG I shot into Adobe Photoshop, used Camera Raw to lower the exposure by two notches, and pulled the highlight and white values to the minimum. Unfortunately, all the Sensor images don't pull back the details of the sky, and they're still dead white.
In addition to Sensor A, some of the leaves hidden in the background highlights have been restored a little. It seems that when shooting, if HDR is not turned on, the details in the highlights in the mobile phone shooting screen can hardly be restored.
Since the details of the sky cannot be pulled back when overexposed, what is the effect of letting the sky be exposed normally and brightening the characters through post-production means? After all, there is an old saying in digital photography called "it is better to owe than to expose", so try it.
I imported the JPG I shot into Adobe Photoshop, and used Camera Raw to increase the exposure by two notches, with a highlight value of -100 and a shadow value of +100. It can be seen that these 4 Sensors can roughly restore the Xinghua students in the case of ensuring that the sky has details, but the picture quality is very touching, accompanied by a lot of brightness, color noise, and there is also a slight color distortion in the shadows.
Judging from the two sets of photos just now, because the size gap between the mobile phone Sensor and the mainstream camera Sensor is too large, if you do not rely on the algorithm, you will be suppressed to death by the physical law of "the bottom is justice", and the recovery adjustment of overexposure or underexposure will cause a greater loss to the picture quality, which is also a reference for students who like to use mobile phone JPG to retouch pictures.
So after turning on the HDR algorithm, how does the picture look and feel? This is related to the HDR algorithm technology of mobile phone manufacturers.
According to the real shooting scene, Sensor A and D restore more realistic and natural, and Sensor B sky color is more colorful. Perhaps weighing different exposure priorities, Sensor C's characters look more flattering, but the sky area is already a bit overexposed.
Xinghua students rushed to the meeting, so the author quickly let him go, after watching the quality of these Sensor and tolerance test performance, what do you want to say? After reading this article, you are welcome to interact together in the comments section.
Fourth, night scene performance
Night scene shooting has always been a nightmare for mobile phones, and the photos taken directly by the small bottom Sensor are not as clear as the eyes can see, and the detail loss is serious. In order to cope with night scenes, some models of Sensor also use RYYB pixel arrangement to increase the amount of light inlet, while adding a powerful night scene algorithm. So turn on the night mode that comes with the device, how do these Sensor night scenes perform? The author fixed these phones to a tripod and enabled the tripod mode of the device to shoot.
Test scenario: The main entrance of the company at night
The principle of night mode is similar to HDR at large light ratios, which is composed of high-speed frames and low-speed frames superimposed to give the finished film a "high" dynamic range.
This "high" dynamic range is reflected in the fact that the highlights are not exposed, and the shadows have details. In the night scene, which is more complex than the environment, see which Sensor has more eye-catching night view capabilities. The author took the highlighted part of the picture, that is, our PConline logo, you can clearly feel that Sensor D's highlight suppression ability is the strongest, followed by Sensor A, while the rest of sensors have overexposed situations, Sensor A night mode of the picture solidity is quite dominant.
Looking at the situation in the dark places, the dark places of Sensor A and C have both a lot of detail and a clear color level. In general, in the balance between the highlights and the dark places of the picture, Sensor A and D have their own advantages, followed by Sensor C, and Sensor B still needs to work.
5. White balance and color
"Straight out" means that the picture has not been modified output, so the white balance and color tendency of the picture determine the taste of "straight out", in this link you can feel the straight out effect of each Sensor.
Test scene: Group studio
Sixth, the focus speed
The last link is to test the focus speed of each Sensor in different light, here the author placed two dolls with different distances, one in the closer place, one in the distance, with 150W film and television lights to control the brightness of the ambient light.
The author first adjusted the brightness of the film and television lights to 100%, and you can see that the focus speed of each Sensor is quite rapid, basically referring to where to hit, without dragging mud and water.
So when the light is poor, what happens to the focus speed of each Sensor? So the author reduced the brightness of the modeling lamp to 1%. The result is obvious, the focus speed of each Sensor has decreased to a certain extent, the best performance is Sensor C, D, the speed drop is not obvious.
VII. Conclusion
Finally, there is the corresponding Sensor model, Sensor A is GN1, Sensor B is S5KHMX, Sensor C is IMX766, and Sensor D is IMX700. Although their parameters are different, the performance of some links is not as large as imagined.
From the entire test results, it can be seen that although there is a parameter gap in the mobile phone Sensor, but it is not magnified 200%-300% to see, in fact, it is not obvious, the comprehensive ability of these Sensors is very good, but in the tolerance of the link, the physical law of "the bottom of the big one crushing people" is still a constraint, and at the same time, the HDR and night mode of these Sensors still have a large gap with what we see with the naked eye.
(iPhone 13 Pro Max main camera with customized IMX703)
In recent years, we can also see that many manufacturers have adopted the solution of custom Sensor, in the case of sensor size can not be significantly increased, focus to optimize pixel arrangement and pixel design, enhance color acquisition, light perception and focus speed, but also to a certain extent to improve the image quality or resolution. Combined with more and more mature algorithms and the optimization of image information links, I believe that mobile phone images in 2022 will have wonderful performances, thank you for reading.