For image capture and creation, SozoTek brings Crystal technology, a new suite of signal processing algorithms that advance the image pipeline in the critical areas of noise and aliasing artifacts. Crystal algorithms, working inside the camera image capture system, add to or replace key image pipeline processes to produce stronger, higher quality images from existing lens and sensor combinations.
Crystal algorithms utilize innovative scientific methods, developed in the SozoTek labs, to remove color aliasing and complex noise artifacts from raw RGB images. Crystal processing generates greater dynamic range and richer colors in all lighting conditions, even low light so common in mobile imaging. The combination of noise removal, pixel correction, edge and color enhancement produces amazing results in cell phone cameras, image sensor modules, DSC, SLR, medical and video applications. This embedded technology is fully scalable and independent of sensor types within the Bayer array configuration.
Remove noise without removing detail Crystal uses simple and unique patented signal processing to isolate noise vs. image detail at the Bayer array. Once the spatial and frequency noise complex is identified from the RGB image, an innovative technique is applied to isolate and remove the noise adaptively, leaving the original image intact. Crystal better identifies and reduces noise at all frequency levels in the image. Noise is reduced without removal of important image detail - especially important in mobile images where image detail is limited.
Generate richer, truer colors Traditional color space transformation, or color correlation of the sensor, is performed to align deviations in sensor sensitivity from the desired color space. Traditional color correlation processes often amplify the present noise, which then is consequently suppressed, creating a color-correct, but blurred image.
Because Crystal so effectively suppresses noise and eliminates chroma aliasing artifacts from the RGB channels, Crystal eliminates the traditional trade-off that occurs between color fidelity and noise amplification. Use of Crystal allows substantially cleaner separation of luminance and chrominance; increasing edge sharpness and color differentials with better white balance. Thus, colors are richer and natural with greater subject and detail clarity may not be possible before.
Crystal easily integrates into today’s embedded systems Crystal algorithms are designed to be economically integrated in existing embedded systems and ISP (image signal processing) pipelines. Several unique algorithms have been created in C and tested in firmware on ARM architecture, and are also applicable to other similar processors. Hardware architecture and models have been created for easy migration and implementation to RTL for IC applications. The HW implementations can be done with synthesizing Crystal IP to image sensors, discrete ISP or can be integrated with applications processor and SOC solutions.
Crystal delivers
High performance - eliminates noise, artifacts and aliasing while enhancing color response
Embedded flexibility - hardware and firmware implementations
Efficient and scalable - any camera resolution, one technology for all ISPs
Economical - appropriate for any product tier, allowing wider BOM vendor choices for lowered system costs
Robust - cleaner RGB data improves every function in the ISP
