I’m filling in at The Verge for 6 weeks. Ask me anything!

2026-07-10 · Source: The Verge · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Computational Photography · Depth: Intermediate, extended

Summary

David Imel, a technology reporter and co-host of the Waveform Podcast, details the evolution of mobile phone photography, tracing its roots to Renaissance art principles like sfumato and chiaroscuro. He explains how the miniaturization of camera sensors, driven by the demand for thinner phones, necessitated computational photography to "cheat physics." Techniques like electronic gain, noise reduction, and sharpening, while improving small sensor performance, often resulted in flat, plasticky images. Google's HDR+, introduced by Mark Levoy around 2016, was a significant advancement, using multi-frame averaging to enhance dynamic range. However, Imel argues that as modern phones now feature larger sensors (e.g., Pixel 6's 1/1.3 inch sensor vs. Pixel 5's 1/2.55 inch sensor), their processing algorithms still over-process images, losing natural depth and contrast. He highlights that raw files from these larger sensors demonstrate superior natural characteristics, suggesting phones should automatically utilize these capabilities. Apple's iPhone 13 "Photographic Styles" offer a promising step towards user-controlled, in-camera processing.

Key takeaway

For product managers and engineers developing smartphone camera algorithms, recognize that current computational photography, designed for small sensors, often over-processes images from newer, larger sensors. You should prioritize algorithms that utilize the natural characteristics of bigger sensors, like improved dynamic range and depth of field, rather than applying aggressive noise reduction and sharpening. Explore user-configurable processing styles, similar to Apple's, to allow users to achieve more natural, artistically contoured photos directly from their devices, avoiding the "plasticky" look.

Key insights

Mobile computational photography, initially a fix for small sensors, now over-processes larger sensors, sacrificing natural image quality and artistic depth.

Principles

• Chiaroscuro and sfumato define photographic depth.
• Miniaturization demands computational compensation.
• Over-processing can flatten natural image aesthetics.

Method

HDR+ captures multiple short exposures, stores them in RAM, aligns frames, and averages pixel colors. This allows gain amplification with reduced noise, enhancing dynamic range.

In practice

• Capture raw images on modern phones.
• Utilize Apple's "Photographic Styles" for custom processing.
• Compare phone JPEGs against raw files.

Topics

• Computational Photography
• Mobile Camera Technology
• Image Processing Algorithms
• Sensor Miniaturization
• HDR+
• Photographic Styles

Best for: Computer Vision Engineer, AI Product Manager, Tech Journalist, Product Manager, Domain Expert

Related on AIssential

• An Extensible and Lightweight Unified Architecture for Demosaicing Pixel-bin Image Sensors — A unified, lightweight architecture with CFA identification improves pixel-bin sensor demosaicing quality and efficiency.
• Adobe’s ‘natural look’ camera app embraces generative AI — Adobe's Indigo app integrates generative AI, powered by Google's Nano Banana, for diverse photo editing and guidance.
• How software learned to make blurry photos clear — Machine learning models, trained on image pairs, reconstruct plausible detail from blurry photos, moving beyond fixed mathematical filters.
• Making the Discrete Continuous: Synthetic RAW Augmentations for Fine-Grained Evaluation of Person Detection Performance in Low Light — Synthetic RAW augmentation can continuously evaluate low-light person detection, overcoming real data sparsity.
• Towards Photorealistic and Efficient Bokeh Rendering via Diffusion Framework — MagicBokeh unifies bokeh rendering and super-resolution using a diffusion framework for photorealistic effects on low-resolution images.
• Seeing Around Corners Using Smartphone-Grade Lidar — Smartphone-grade lidar, enhanced by algorithms, can perform non-line-of-sight imaging, making advanced sensing accessible.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Verge.