Image Sharpening

(1) Supports a wide range of blurring types

Various factors, such as camera shake or focus issues, can cause blurring in camera images. There are various technologies for eliminating blurring, but one of the keys to improving image quality is selecting the right correction method based on the cause of the blurring. For example, in the case of motion blur (caused when the subject is moving), even if image enlargement blur correction (generally referred to as “super-resolution”) is applied, it won’t help reduce the blurring. That’s why “Morpho Deep Deblur” offers four different blur correction AI models, to perform the type of blur correction that is most effective given the cause of the blurring. This makes it possible to achieve accurate image quality correction.

Morpho also offers blur correction AI model customization services based on the processing requirements of individual customers.

＜AI Model Selection＞

– Defocus blur correction AI model

This AI model is used to make corrections to photographs which were taken out-of-focus. This type of correction can be applied when you realize after the fact that the camera was focused when a photo was taken. It sharpens blurry images and enables you to keep a crisp, clear image of fleeting moments.

The underlying mechanism of defocus blur is that in order to increase image brightness when taking photographs, a camera’s diaphragm is opened wide (to let in a large amount of light), which reduces the depth of field. This increases the probability of partial defocus blur.

– Motion blur correction AI model

Image brightness can also be increased when taking a photo by lowering the shutter speed. This can result in motion blurring due to the movement of the photo subject. This AI correction model reduces motion blur. Furthermore, “Morpho Deep Deblur” can be used in conjunction with “Morpho Motion Sensor™” technology (sold separately) for detecting the movement vectors of moving subjects to determine if the speed falls within correctable bounds.

– Image enlargement blur correction AI model

This AI model is used to correct for reductions in resolution caused by enlarging an image. It is generally referred to as “super-resolution” technology. It can be used for image quality correction in situations where bandwidths are limited, such as streaming, by compressing image sizes before transmitting them, enlarging them after receiving them at the other end, and then applying this AI model.

It can also correct for aliasing and loss of detail in images that were photographed using a digital zoom, producing a smooth, detailed image.

– Distortion correction-induced blur correction AI model

This AI model is used to correct for reductions in resolution at the edges of images caused by applying distortion correction.

(2) Uses the “SoftNeuro®” AI inference engine

Morpho’s high-speed AI inference engine “SoftNeuro” is built into the “Morpho Deep Deblur”. It supports various hardware (CPUs, GPUs, and AI accelerators) and fully leverages hardware resources to perform object detection processing quickly, while consuming little power. It can be used on x86/x64/ARM CPUs, GPUs, and AI accelerators.

– AI Inference Engine “SoftNeuro”: https://www.morphoinc.com/en/technology/sie

(1) Before and after image enlargement blur correction

(2) Before and after defocus blur correction

“Morpho Deep Deblur” can be used in many fields, including the automotive and smartphone/embedded computer camera fields.

●Automotive field

In recent years, there has been an increasingly prominent need within the market for image enhancement software technologies for taking low-resolution video images from on-board cameras and displaying them as clear images on large display monitors (HMI). This demand has been driven by major recent advances in HMIs. Increases in HMI resolutions are prompting demands for higher on-board camera resolutions. However, the use of high-resolution cameras significantly increases vehicle costs, so there is a need for software technologies that can make corrections to low-resolution video from on-board cameras and present clear images on HMIs.

“Morpho Deep Deblur” can be combined with HMIs as a hub that uses AI to improve input image quality. This allows clear images to be displayed on large monitors, even if the source images are low resolution.

Furthermore, “Morpho Deep Deblur” has been added as a standard AI technology to “Morpho Automotive Suite™”, the AI solution package for the automotive field released in May 2023. In addition to supplying individual AI technologies, Morpho is also helping solve problems faced in the automotive field and adding value to automotive products by offering additional AI learning fitted to customer needs and by providing proposals such as product customization proposals for specific customer applications.

●Smartphone/computer/digital camera field

“Morpho Deep Deblur” can be applied to address issues with lower resolutions caused by using digital zoom functions on smartphones or digital cameras or by enlarging photographs. There are many super-resolution technologies based on signal processing, but “Morpho Deep Deblur” uses AI to go beyond what is feasible using a signal processing approach and achieve even greater resolution improvements.

*The product name “Morpho Deep Deblur” is a tentative name and is subject to change.