18514252. ADAPTIVE DEFORMABLE KERNEL PREDICTION NETWORK FOR IMAGE DE-NOISING simplified abstract (Intel Corporation)

ADAPTIVE DEFORMABLE KERNEL PREDICTION NETWORK FOR IMAGE DE-NOISING

Organization Name

Intel Corporation

Inventor(s)

Anbang Yao of Beijing 11 (CN)

Ming Lu of Beijing 11 (CN)

Yikai Wang of Beijing (CN)

Xiaoming Chen of Shanghai 31 (CN)

Junjie Huang of Shenzhen (CN)

Tao Lv of Shanghai (CN)

Yuanke Luo of Shanghai (CN)

Yi Yang of Shanghai 31 (CN)

Feng Chen of Shanghai 31 (CN)

Zhiming Wang of Shanghai 31 (CN)

Zhiqiao Zheng of Shenzhen (CN)

Shandong Wang of Beijing 11 (CN)

ADAPTIVE DEFORMABLE KERNEL PREDICTION NETWORK FOR IMAGE DE-NOISING - A simplified explanation of the abstract

This abstract first appeared for US patent application 18514252 titled 'ADAPTIVE DEFORMABLE KERNEL PREDICTION NETWORK FOR IMAGE DE-NOISING

Simplified Explanation

The adaptive deformable kernel prediction network for image de-noising is a method that uses a convolutional neural network to filter pixels in an image to obtain a de-noised pixel.

• Convolutional kernel is generated for each pixel in the image.
• Offsets are generated for each pixel to indicate deviations from the pixel position.
• Deviated pixel positions are determined based on the pixel position and offsets.
• Filtering is done using the convolutional kernel and pixel values of deviated pixel positions to obtain a de-noised pixel.

Potential Applications

This technology can be applied in various fields such as:

• Image processing
• Computer vision
• Medical imaging
• Surveillance systems

Problems Solved

This technology helps in:

• Removing noise from images
• Enhancing image quality
• Improving accuracy in image analysis

Benefits

The benefits of this technology include:

• Improved image clarity
• Enhanced image recognition
• Better performance in image-based tasks

Potential Commercial Applications

This technology can be commercially applied in:

• Photography software
• Security systems
• Medical imaging devices

Possible Prior Art

One possible prior art for this technology could be traditional image de-noising algorithms that use filters and convolution operations to remove noise from images.

What are the specific kernel values used in the convolutional kernel for image de-noising?

The specific kernel values used in the convolutional kernel for image de-noising are generated for each pixel in the image.

How does the method determine the deviated pixel positions based on the pixel position and offsets?

The method determines the deviated pixel positions by using the pixel position of the pixel and the offsets generated for each pixel.

Original Abstract Submitted

Embodiments are generally directed to an adaptive deformable kernel prediction network for image de-noising. An embodiment of a method for de-noising an image by a convolutional neural network implemented on a compute engine, the image including a plurality of pixels, the method comprising: for each of the plurality of pixels of the image, generating a convolutional kernel having a plurality of kernel values for the pixel; generating a plurality of offsets for the pixel respectively corresponding to the plurality of kernel values, each of the plurality of offsets to indicate a deviation from a pixel position of the pixel; determining a plurality of deviated pixel positions based on the pixel position of the pixel and the plurality of offsets; and filtering the pixel with the convolutional kernel and pixel values of the plurality of deviated pixel positions to obtain a de-noised pixel.