Technology / Product Information

<Innovative Unique Technology>
Real-time Blind Deconvolution

We are developing various products centered on our completely original PSF blind real-time deconvolution technology RING (Restoration with INtegral Geometry technology) that is not affected by lens characteristics.

RING has three technological advantages: sharpening, ultra-high-speed processing, and low power consumption.

With the sharpening performance, blurs are improved up to 2 times.

With the ultra-high-speed processing performance, 4K images are processed in 9msec per frame. This is up to about 1 million times faster than that of the Modified-Lucy method, which is the conventional blind deconvolution technology.

With the low power consumption performance, images are processed at 0.35wh at full VGA30FPS operation.

Software can exceed the limits of hardware and can apply to all areas that utilize digital optical systems.

RING solves problems such as insufficient performance of equipment, transmission/reception of large-capacity images, and insufficient image quality, which have been difficult to solve.

<Forms of Product Provision>

Currently, we are developing three products: desktop analysis software for research and development, embedded API / original codec, and cloud service.

We can provide the optimum combination according to the customer's problem.

Please feel free to use the inquiry form below to request materials or make inquiries regarding our products.

<Performance and Features of RING>

■Image Quality (Image Sharpening Effect)

With the sharpening process, you can mainly obtain the effects of (1) edge sharpening, (2) contrast improvement, and (3) resolution improvement.

With the dramatic improvement in image quality, the appearance as well as the profile data will improve.

Therefore, we can expect improvements in the performance of various judgments and computing processes (machine learning, AI, recognition system, etc).

Left: Original image (image obtained from a camera)

Right: Image after sharpening process by RING (image after processing the original image with RING)

■Ultra-High-Speed Processing Performance

Ultra-high-speed and lightweight processing are available with a normal general-purpose PC. This technology completely surpasses conventional deconvolution technology such as the RL method.

Since RING allows to obtain real-time images that were previously difficult to utilize, it can operate using any device, from high-speed cameras to ordinary cameras.

Blind deconvolution processing performance of RING (@PassMark = 19364).

■ Low Power Consumption

Due to the surprisingly low amount of calculation, the load on the CPU / GPU is extremely reduced. RING operates with low power consumption due to improved energy efficiency. It works lightly even on low-performance computers.

As a result, not only the electricity bill and CO2 will reduce, but also the cloud server specifications will downsize, and the battery life and service life of mobile devices will improve.

■Application Examples:
Situations where RING demonstrates its strengths.

(1) Situations where the resolution is too high and blurring is noticeable.

e.g.) miniaturized and high-pixel surveillance camera

(2) Situations where infrared wavelength (IR) camera blur is noticeable.

e.g.) night-vision camera, satellite observation, non-destructive inspection

(3) Situations where you want to reduce optical damage to the sample.

e.g.) biological microscope

(4) Situations where blurring is noticeable when the depth of field is deep.

e.g.) inspection equipment, biological microscope, metallurgical microscope

(5) Situations where the size and the weight of the lens are causing problems.

e.g.) small cameras, drones, satellite observations

(6) Situations where the invalid magnification caused by optical zoom is conspicuous.

e.g.) enlarged images

(7) Situations where defocus blur is conspicuous.

e.g.) Phase-contrast microscope

(8) Situations where information extraction throughput is poor due to the combined reasons of the resolution, SN value, and computational complexity.

e.g.) replacement of Unsharp Mask, replacement of existing deconvolution

■Usage Cases

If the target image itself has a research issue, we recommend the A pattern in the figure below.

We will make optimal proposals based on the knowledge cultivated in many industrial and academic fields.

The above are only a part of the cases that we have demonstrated, improved, and solved.

We will make the best proposals to our customers based on our abundant achievements, so please feel free to contact us and request product materials.