SozoTek has distilled years of research to create Crystal, breakthrough embedded science that substantially improves image pipeline processes and RAW Bayer data at the image sensor. Crystal algorithms dramatically remove multiple noise levels and aliasing artifacts frequently found in mobile embedded imaging systems.
Why Crystal™? With accelerating numbers of digital photos captured in available light, driven by the prevalent use of camera phones, traditional RAW image pipeline processes are being challenged to generate high quality images when light intensity and color temperatures vary greatly at image capture. Accentuating the problem in cell phone images is the fact that the actual image data is often severely limited or compromised by lens or sensor inadequacies, in addition to being buried in noise. As phone cameras increase in sensor resolution, but light remains limited from lens size and cost restrictions, noise problems accelerate. Increasing the number of pixels alone is not a viable long term solution when dealing with inherent noise issues.
Crystal Signal Processing SozoTek Crystal algorithms operate on the raw RGB Bayer data, performing smart interpolation and digital signal processing within the embedded imager. The heart of process involves analyzing each pixel output from the image sensor A/D and the interaction of neighboring RGB pixels at multiple frequencies and spatial relationships. From Crystal’s advanced luminance and chroma interaction analysis a composite noise and chroma aliasing map is extracted. The isolation of this complex map is then utilized for noise cancellation, effectively removing unwanted noise from the image and restoring the original image content. Each RGB band is rendered robust and clean, providing far more dynamic range for subject color correction and edge enhancement to suite a variety of customer scene preferences.
Most interpolation techniques are unable to separate enough noise from the image system for subsequent sharp color corrections, resulting in images that appear blurred or lacking in beauty depth. Crystal’s remarkable success in noise and aliasing removal is a result of the algorithm’s high accuracy in complex aliasing and noise measurements and its ability to remove noise energy adaptively. Colors and edges within the original image are thus hardened and preserved to reproduce DSC-quality images.
SozoTek Crystal algorithms, patent-pending, are deployable in firmware and hardware architectures. They can be implemented in the existing ISP to replace or augment existing pipeline processes as required for each system architecture, or implemented within the image sensor, discrete IC, or integrated with the processor SoC.
SozoTek signal processing algorithms are inherently scalable and adaptable to a wide range of resolutions independent of sensor-vendor type and variations in Bayer-array systems. SozoTek flexible embedded architectures provide freedom of design, and economies of scale, for multiple tiers of products, and allows sourcing flexibility from multiple image sensor or IC vendors.
