True Colors through Multispectral Imaging AWB
Modern smartphone cameras are essentially color-blind.
Traditional Auto White Balance (AWB) algorithms, based only on RGB sensors, often misjudge the correct white point and rely on manually tuned, ad-hoc rule sets. With only three color channels available, these methods face ambiguity in illuminant estimation, resulting in poor color fidelity — especially in scenes with multiple light sources or complex lighting environments.
Spectricity’s multispectral AI-powered AWB changes that.
By combining advanced deep-learning algorithms with rich multispectral data, our system detects and characterizes illuminants through their unique spectral signatures. This enables precise spatial white-point prediction and delivers true-to-life colors, natural skin tones, and consistent performance across diverse lighting conditions — from indoor LED to outdoor daylight.
Traditional AWB: RGB
- White point ambiguity due to only 3 color channels
- Manually tuned extensive ad-hoc rule set
Spectricity AWB: Multispectral AI
- Improved spatial white point prediction due to advanced AI deep learning and rich multispectral data
Advancing the Field: Local AWB Powered by Multispectral Imaging
At Spectricity, we’re redefining what’s possible in mobile imaging. Building on our leadership in multispectral imaging technology, we’re taking Automatic White Balancing (AWB) to the next level, with local AWB.
Traditional AWB solutions assume a single, uniform light source. But real-world scenes are far more complex, think of an indoor shot illuminated by warm indoor lighting and cool daylight streaming through a window. Conventional RGB-based algorithms simply can’t capture these variations accurately, leading to inconsistent colors across the image.
Our multispectral imaging technology changes that. By capturing richer spectral information at the pixel level, Spectricity enables precise, localized white balance correction, delivering true-to-life colors in every part of the scene.
With local AWB, Spectricity sets a new benchmark for next-generation, color-intelligent imaging systems, bringing professional-grade color accuracy to every device.
Request Our MIMI Dataset
Multi-Illuminant Multispectral-Imaging
Spectricity is committed to advancing the field of AWB. In line with this mission, we are making the Multi-Illuminant Multispectral-Imaging (MIMI) dataset publicly accessible.
The MIMI dataset includes both standard RGB and MSI images from nearly 800 scenes, making it an ideal resource for developing white balancing algorithms for multi-illuminant environments.
More details about the dataset are available on its GitHub page.
To gain access to the dataset, simply complete the form on the right.
*Important: This dataset is available for research, academic, and non-commercial purposes. Please use a valid organizational or company email address and provide a brief description of your intended use. All submissions will be reviewed manually. Requests with incomplete or unverifiable information may be declined.
Read more about the underlying chip technology enabling high-volume spectral imaging.