The moon in domestic mobile phones looks more perfect.

On the night of the Lantern Festival on the 15th day of the first lunar month, a rare "red moon" will hang in the sky. For the public in China, there will only be three opportunities to see the "red moon" on the Lantern Festival in the 21st century, namely in 2007, 2026, and 2072.

In the past five or six years, a peculiar astronomical - like phenomenon has emerged in the Chinese smartphone industry: on the nights of a full moon or a supermoon, tens of thousands of close - up photos of the moon appear on social media.

In these photos, the moon's texture is clear and craters are visible. In contrast, what iPhone users can usually show off is often an overexposed white dot. This huge difference in experience has made photographing the moon a regular feature at domestic mobile phone press conferences, used to prove that their phones' long - focal - length night - shooting capabilities are far ahead.

The imaging industry generally believes that an important fork in this technological path occurred in 2019. At that time, Huawei's P30 series was promoting periscope - style long - focal - length lenses and found a scenario where consumers could easily see the hardware difference at a glance: photographing the moon. Since then, this function has become a standard feature of high - end domestic mobile phone models.

Users suddenly found that the moon in their phones looked much better than the one they saw with their own eyes. At the same time, users' doubts about "photoshopping the moon" have never stopped. Some phones have photographed two moons, some have shown the moon piercing through branches in the foreground, and some have even misidentified streetlights as the moon - after a few seconds of algorithm processing, adding crater textures to the streetlights...

For mobile phone manufacturers, the moon is a standardized object that most users can easily photograph. It is the best billboard to prove hardware capabilities, especially long - focal - length anti - shake and large apertures. Each manufacturer racks its brains on how to more accurately identify the moon, how to make the moon and ground objects (such as city landmarks) have a perfect group photo, and what magnification ratio of the moon looks more natural. Some manufacturers even have dedicated teams internally to maintain the moon - related algorithms.

Taking the Perfect Moon Photo

On January 3, 2026, the first supermoon of the year appeared. That night, a photography enthusiast drove nearly 400 kilometers but couldn't get out of the rainy and cloudy weather. Finally, he took a perfect photo of the moon with his son's OPPO phone and posted it on his WeChat Moments. He commented that domestic mobile phones have the ability to "pierce through the clouds" and make the moon look brighter.

"It's very difficult to take such clear close - up photos of the moon relying solely on the existing hardware of the phone without algorithms," said Hou Weilong, a senior imaging expert at Honor, without avoiding the topic of algorithms. He explained that the moon is the second scenario after human faces where generative AI is widely used.

A photographer explained to Economic Observer that only cameras with a hardware combination of a long - focal - length lens of over 400mm and a high - dynamic - range photosensitive element can capture the shadow levels and crater outlines on the lunar surface. The long - focal - length photosensitive elements of mobile phones are usually only 1/1.3 inches or even smaller, and the amount of incoming light is far less than that of APS - C cameras. After some mobile phones identify the moon, they use generative algorithms to directly add moon textures.

Due to tidal locking, the moon always has only one side facing the Earth. This means that whether people are in Beijing or New York, the texture of the moon they see is the same.

"As long as the phone identifies it as the moon, mobile phone manufacturers have an imaging algorithm that takes into account the laws of lunar phase changes. The algorithm will generate and enhance based on the learned textures," Hou Weilong gave an example. It's like facial beautification. Although people have different faces, pores, skin textures, and eyebrow directions are common. The algorithm enhances details on a blurry basis but won't replace one person's face with another's.

There has always been a controversy about whether imaging algorithms are a form of fraud. The aforementioned photography enthusiast doesn't oppose the moon - related algorithms: everyone sees the same side of the moon, and mobile phone manufacturers use hardware + post - processing algorithms to directly give users a moon closer to the "standard answer." However, he differentiates between "photography that records real light" and "images calculated by algorithms."

Embarrassing Misidentifications

To take clearer photos of the moon, domestic mobile phones have continuously iterated on long - focal - length hardware and imaging algorithms. However, in the past six years, many users on social media have reported that their phones have misidentified streetlights and big cakes as the moon, or there have been bugs where the algorithm rigidly wipes out the foreground branches, leaving a floating fake moon.

How to avoid taking fake moon photos? Hou Weilong explained that in the early days, relying solely on image feature recognition made it easy to mistake streetlights for the moon. The team he is on introduced a multiple verification mechanism: first, before detecting the moon, the phone's algorithm reduces the exposure of the picture to confirm the moon's surface texture; second, before deciding whether to enhance the moon, the phone first calls on GPS location, system time, and gyroscope data. The system then calculates whether there is a moon in the sky in the direction the user's phone is facing at that moment. If the direction is wrong, even if there is a bright circular object (such as a streetlight) in the picture, the algorithm won't intervene for enhancement.

On the matter of photographing the moon, different manufacturers have developed different approaches.

A person in charge of imaging at a leading mobile phone manufacturer revealed that manufacturers like vivo have a separate "Super Moon" or landscape mode, which magnifies the moon by more than 20% to create a visual impact beyond what the naked eye can see. Most manufacturers usually only do a slight magnification of 10% - 15%, emphasizing more on maintaining the real relationship between the foreground and the background.

In a scenario where the moon is bright and the stars are few, mobile phone imaging algorithms can already enable most users to take close - up photos of the moon. What's more difficult is to handle some non - standard variables, such as clouds and branches in the foreground that block the view.

Hou Weilong explained that if the clouds are very thick and completely cover the moon, the algorithm can't recognize it, and the phone will take it as an ordinary night scene. The most difficult situation to handle is thin clouds: the clouds are semi - transparent, and the moon's texture becomes weaker. If the algorithm forcibly enhances the moon's texture, it will look like the moon is "stuck" in front of the clouds. This is a technical challenge.

There is a small team within Honor dedicated to researching algorithms related to the moon. Previously, the team did basic work such as collecting moon - related materials as a training database. Hou Weilong revealed that in the past two years, Honor's focus has been on "moon - landmark group photos," that is, taking good photos of the moon and landmark buildings. "This is very difficult because the brightness contrast between the moon and objects on the ground at night is very large, and the dynamic range exceeds the coverage of conventional cameras."

Computational Photography Offers Higher Cost - Effectiveness

A person related to the algorithm of an action camera explained that many high - end flagship mobile phones feature night - shooting capabilities, but taking good photos of the moon is not the most important manifestation of night - shooting ability. From a business perspective, photographing the moon should have been a discarded scenario. Compared with the high - frequency portrait and landscape photography, users rarely take photos of the moon, perhaps only once or twice a month. "R & D resources are limited, and the priority of the moon is much lower than that of portraits."

Hou Weilong defined photographing the moon as a low - frequency but high - value essential need. In the context of Chinese culture, the moon is a totem of reunion and longing. "When others can take photos of the moon but you can't, users will think there is something wrong with the phone." This obvious difference is enough to be a reason for users to change their phones.

Therefore, domestic manufacturers have turned a low - frequency scenario into a standard feature of imaging flagships. To avoid errors when the algorithm "creates something out of nothing," manufacturers must invest in physical long - focal - length hardware.

The cost of hardware is high. The person related to the action camera algorithm often receives feedback from users asking if the next new product can have a 1 - inch large - sized sensor (CMOS sensor), believing that "a larger sensor is always better." However, in recent years, many manufacturers have stopped insisting on the 1 - inch sensor and instead have moved towards 1/1.3 inches or 1/1.4 inches.

He explained that not using the 1 - inch sensor is not only because a single sensor is expensive. After using it, the phone needs to solve problems such as the size of the module, power consumption, and heat dissipation design, which are all real costs. Currently, the mainstream approach in the industry is to use a 1/1.3 - inch sensor and, through algorithms such as multi - frame noise reduction, handle night - scene noise better than full - frame cameras. He believes that this is a more cost - effective path that computational photography can achieve.

In contrast, Apple is relatively conservative in imaging algorithms. The person related to the action camera algorithm interpreted domestic manufacturers' bet on computational photography as a strategy similar to Tian Ji's horse - racing.

He believes that in the field of imaging, Apple's strength lies in video. Video places more demands on the core hardware capabilities of the chip (SOC), and Apple's chips are self - developed, allowing it to design the ISP (Image Signal Processor) according to its own needs. In contrast, most Android phones use general - purpose chips such as Qualcomm's, and they can't be deeply customized like Apple's. Currently, no manufacturer can introduce too many algorithms in real - time in 30 - frame - per - second or 60 - frame - per - second videos.

Facing the sales volume of tens of millions of a single iPhone product, domestic flagships must focus on scenarios such as portrait beautification and long - focal - length moon photography to compete for female users or customers in specific markets.

Hardware has physical limitations, while algorithms have almost no limit with the support of AI and are more cost - effective. How do mobile phone manufacturers choose between the two? Hou Weilong said that flagship products don't make a choice. Hardware is the foundation, and algorithms are the enhancement. For mid - and low - end products, manufacturers use algorithms to make up for the lack of hardware.

When the moon has been photographed clearly enough, even to the point of being boring, the industry has started to look for the next exciting point.

"The industry is shifting from 'taking good photos' to 'taking good photos of what,'" Hou Weilong believes. Nowadays, young people no longer regard clarity as the only goal of photography. Instead, they pay more attention to personal expression and even pursue the blurry atmosphere of CCD cameras. In the future, mobile phones will not only be cameras. With the support of AI, they can become an AI photography advisor, for example, telling users "where to take a photo of the moon passing through the Spring Bamboo Tower" or "how to compose a better picture in this light."

The person related to the action camera algorithm has also observed similar changes. Some camera manufacturers have started to launch physical handle sets, which may not necessarily improve the image quality but provide a sense of ceremony and emotional value in shooting, such as making the moon look rounder and brighter. Domestic mobile phone manufacturers are competing to better understand the person standing under the moonlight taking photos.

This article is from the WeChat official account "Economic Observer". Author: Chen Yueqin. Republished with authorization from 36Kr.