Back to results
US 2022/0036510 A1Published application
Method and image processing device for generating image filter
AssigneeNovatek Microelectronics Corp.
InventorsXin Huang, Fan Wang, Xi Tian
Life of this patent
Estimating the term…
Abstract
A method and an image processing device for generating an image filter are proposed. The method includes the following steps. A first image is received and inputted into a pre-trained convolutional neural network (CNN) model, and the image filter corresponding to the first image is obtained based on an output result of the pre-trained CNN model.
Claims20
A method for generating an image filter, comprising:
- receiving a first image;
- receiving a second image succeeding the first image in an image sequence;
- inputting the first image into a pre-trained convolutional neural network (CNN) model, wherein the step of inputting the first image into the pre-trained CNN model comprises:
- splitting the first image into a plurality of color channels to generate a plurality of first split images respectively corresponding to the color channels;
- splitting the second image into the color channels to generate a plurality of second split images respectively corresponding to the color channels;
- computing a plurality of motion vectors based on the first split images and the second split images;
- and
- inputting the first split images, the second split images, and the motion vectors into the pre-trained CNN model;
- and
- obtaining the image filter corresponding to the first image based on an output result of the pre-trained CNN model.
7 more elements
(canceled)
(canceled)
An image processing device comprising:
- a memory circuit, configured to store a pre-trained convolutional neural network (CNN) model;
- a processing circuit, configured to receive a first image, receive a second image succeeding the first image in an image sequence, input the first image into the pre-trained CNN model, and obtain the image filter corresponding to the first image based on an output result of the pre-trained CNN model,
- wherein the processing circuit splits the first image into a plurality of color channels to generate a plurality of first split images respectively corresponding to the color channels, splits the second image into the color channels to generate a plurality of second split images respectively corresponding to the color channels, computes a plurality of motion vectors based on the first split images and the second split images, and inputs the first split images, the second split images, and the motion vectors into the pre-trained CNN model.
(canceled)
(canceled)
Dependencies are inferred from the claim text.