While AI imaging offers immense potential, it must be used judiciously.

2024 marks a pivotal year for AI hardware innovation.

In the AI realm, 'focusing on applications rather than specifications' has become a widely held belief. On one hand, hardware advancements are integrating AI capabilities, leading to an explosion in AI-enabled devices across various industries, including mobile phones, PCs, home appliances, automobiles, cleaning devices, headphones, cameras, and storage solutions. On the other hand, AI is profoundly transforming software, with native AI applications like Wenxiaoyan and Doubao gaining widespread popularity, while search engines, input methods, browsers, payment systems, office software, and e-commerce platforms are undergoing AI-driven reinvention.

AI has emerged as a 'magic wand' for technological innovation, and this is merely the beginning.

In December, we will launch the 'Leitech·Annual' special, featuring 'Focus on 2024' to systematically showcase notable companies, products, technologies, and individuals in the tech industry for the year. Additionally, 'Outlook 2025' will preview anticipated products and technologies in the tech industry for 2025, honoring innovation, documenting our era, and inspiring the future.

Please subscribe and follow us for updates.

Over the past 24 years, the concept of computational photography has gained traction, with imaging becoming the primary battleground for AI applications. Whether it's integrating AI technology into native cameras for enhancement or popular AI-powered apps like AI removal, these innovations have opened new avenues for imaging. However, does the advent of AI imaging truly benefit photographers?

AI Imaging Makes Great Photos Easily Accessible

Early AI photography applications primarily focused on the photo-taking process, with smartphones utilizing deep learning and neural networks to intelligently process images. For instance, when capturing a night scene, the phone automatically adjusts parameters like exposure, contrast, ISO, and color saturation to produce a visually appealing photo. For portraits, it optimizes skin color and texture to enhance naturalness.

Ideally, this sounds perfect. In reality, however, AI often interferes, leading to issues like excessive beautification, unrealistic brightness enhancements, and color casts. While the results may appear better, they often exhibit a distinct 'digital flavor,' differing significantly from what our eyes perceive, prompting many users to disable AI imaging shortly after purchase.

Today, AI imaging on mobile phones has made remarkable progress, with algorithms becoming so seamless that average users barely notice them. Furthermore, it introduces a suite of AI post-processing functions, such as AI clutter removal, AI crowd removal, AI reflection removal, AI recomposition, and AI stylized learning, significantly boosting photo output rates and post-processing speeds.

(Image source: OPPO official)

For example, OPPO's Hasselblad portrait mode integrates AI to meticulously process subjects' fine hairs and fabric textures, enhancing the phone's precision in capturing details. Another instance is Huawei's XD Motion engine, which employs AI to synthesize multiple images, enabling high-speed burst shooting.

While these operations may seem trivial to seasoned photographers, they significantly boost photo output rates and spark a passion for photography among beginners.

Take Capture One, a software I frequently use, as an example. It's a professional post-processing tool offering extensive parameter adjustment capabilities. However, beginners often spend considerable time learning theoretical knowledge and practicing post-processing.

Starting in 2023, Capture One incorporated various AI functions. Initially, it added simple features like automatic white balance and exposure adjustment, already available on mobile phones. With the maturity of large models, it gradually introduced functionalities like automatic 'dust removal,' 'keystone correction,' and 'HDR adjustment.' While these might not be highly technical, they indeed save considerable time.

(Image source: Leitech illustration)

Last November, Capture One introduced 'Smart Color Matching Adjustment.' By selecting any image online as a reference, the software automatically applies its color, exposure, dynamic range, and other parameters to the desired image, simplifying complex post-processing steps and enriching creative possibilities.

(Image source: Leitech illustration)

Similarly, Adobe's Photoshop and Lightroom have incorporated numerous AI functions to expedite photographers' workflows.

But is this development unalloyedly positive?

Is AI Imaging a Double-Edged Sword?

Historically, the mobile photography competition revolved around image sensors, lens modules, color reproduction, and imaging quality, while traditional photography aimed to instantly replicate the real world.

As AI large models gained processing capabilities, smartphone manufacturers believed AI could enhance photo quality independently of hardware, relying solely on computations to alter original image data.

Current AI imaging can be likened to sitting in front of an AI image generator, where inputs dictate the desired output. In mobile photography, built-in algorithms issue instructions, and AI regenerates an image accordingly.

This is evident in mobile phone cameras' AI telephoto feature. Flagship smartphones now advertise AI telephoto technology, using algorithms to restore missing image details.

However, AI is not infallible and can make errors. Leitech previously reviewed four flagship imaging devices for their telephoto capabilities. At 60x zoom, all phones impressed with image quality but exhibited 'void filling,' making images somewhat artificial.

(Image source: Leitech illustration)

Moreover, the prevalence of AI technology may lead to photographic homogenization as similar algorithms produce comparable image effects. Many software now offers color matching and style imitation, blurring the lines between facilitation and restriction. If everyone abuses AI, what remains of photography's essence?

Even more alarming is AI's image generation capability, nearing the point of visual indistinguishability. With proper training, it can create realistic portraits, landscapes, or artworks, potentially devastating the photography industry if misused long-term.

AI Is Redefining Photography

AI-infused mobile phone cameras have indeed lowered the bar for capturing great photos, enabling more people to cherish memories through daily phone snapshots. While they might not understand photography intricacies, flipping through their albums evokes cherished memories. From this perspective, AI's impact on mobile photography is positive and worth promoting.

In AI imaging, powerful hardware underpins algorithmic efficacy. The optical solution comprising sensors and lenses determines a phone's basic capture capabilities, forming the foundation of everything. Without optics, AI imaging degrades into AI generation.

(Image source: Leitech production)

AI and optical imaging should complement each other. As the era of AI algorithm-driven imaging approaches, manufacturers may reverse-engineer imaging chips and sensors based on algorithmic needs, fostering hardware, sensors, and AI algorithm synergy. This will elevate imaging power consumption and effects, further narrowing the gap with mirrorless and DSLR cameras.

AI simplifies and enriches mobile photography, making professional-level photography accessible to all. This is AI imaging's core value.

If AI can be locally trained based on user behavior to create a 'customizable camera,' offering a unique experience per phone, it heralds a promising future. For example, if a user prefers warm-toned photos, the camera could default to a warmer color temperature. If they delete many darker exposure photos from a set, the phone's imaging style might brighten. If they frequently shoot ultra-wide-angle landscapes, the camera app could default to vibrant ultra-wide-angle mode. Over time, users obtain a 'camera' tailored to their preferences.

This is the transformative potential of AI imaging.