IP Library Granted Patent US 11,631,199
Granted Patent B2
US 11,631,199 · App. 16/636,669 · Granted Apr 18, 2023

Image filtering apparatus, image decoding apparatus, and image coding apparatus

Inventors: Tomohiro Ikai (Sakai, JP); Tomoyuki Yamamoto (Sakai, JP); Norio Itoh (Sakai, JP); Yasuaki Tokumo (Sakai, JP)
Assignee: SHARP KABUSHIKI KAISHA
G06T9/002G06N3/045G06T3/4046H04N19/105H04N19/117H04N19/159H04N19/182H04N19/186
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 11,631,199
App. No.
16/636,669
Granted
Apr 18, 2023
Kind
B2
Abstract

To apply a filter to input image data in accordance with an image characteristic. A CNN filter includes a neural network configured to receive an input of one or multiple first type input image data and one or multiple second type input image data, and output one or multiple first type output image data, the one or multiple first type input image data each having a pixel value of a luminance or chrominance, the one or multiple second type input image data each having a pixel value of a value corresponding to a reference parameter for generating a prediction image and a differential image, the one or multiple first type output image data each having a pixel value of a luminance or chrominance.

Claims (28)

1. An image filtering apparatus comprising:

a neural network configured to input (i) one or multiple first type input image data and (ii) one or multiple second type input image data, and output one or multiple first type output image data, wherein

the one or multiple first type input image data each have a pixel value of a luminance or chrominance, the one or multiple second type input image data each indicate a pixel value of a value of a reference parameter used for generating a prediction image or a differential image, and the one or multiple first type output image data each have a pixel value of a luminance or chrominance, and

the second type input image data, derived for a unit of region, is inputted to the neural network as one of image channels.

2. The image filtering apparatus according to claim 1 , further comprising:

a parameter determination unit configured to update a neural network parameter to be used by the neural network.

3. The image filtering apparatus according to claim 1 , wherein

the reference parameter is a quantization parameter in an image on which the image filtering apparatus acts.

4. The image filtering apparatus according to claim 1 ,

wherein the reference parameter is a parameter indicating types of intra prediction and inter prediction in an image on which the image filtering apparatus acts.

5. The image filtering apparatus according to claim 1 ,

wherein the reference parameter is a parameter indicating an intra prediction direction (intra prediction mode) in an image on which the image filtering apparatus acts.

6. The image filtering apparatus according to claim 1 , wherein the reference parameter includes a parameter indicating a partition split depth in an image on which the image filtering apparatus acts.

7. The image filtering apparatus according to claim 1 ,

wherein the reference parameter includes a parameter indicating a size of a partition in an image on which the image filtering apparatus acts.

8. The image filtering apparatus according to claim 1 , further comprising:

a second neural network of which an output image is the first type input image data input to the neural network.

9. The image filtering apparatus according to claim 1 ,

wherein the neural network is configured to input (i) the first type input image data indicating pixel values of a first chrominance and a second chrominance and (ii) the second type input image data, and

the neural network is configured to output the first type output image data indicating pixel values of the first chrominance and the second chrominance.

10. The image filtering apparatus according to claim 1 ,

wherein the neural network includes:

a unit configured to input (i) first type input image data of the one or multiple first type input image data having a pixel value of a luminance and (ii) second type input image data of the one or multiple second type input image data, and to output first type output image data having a pixel value of a luminance; and

a unit configured to input (i) first input image data of the one or multiple first type input image data having pixel values of a first chrominance and a second chrominance and (ii) second type input image data of the one or multiple second type input image data, and to output first type output image data having pixel values of a first chrominance and a second chrominance.

11. An image decoding apparatus for decoding an image, the image decoding apparatus comprising:

the image filtering apparatus according to claim 1 as a filter configured to act on a decoded image.

12. An image coding apparatus for coding an image, the image coding apparatus comprising:

the image filtering apparatus according to claim 1 as a filter configured to act on a local decoded image.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Feb 5, 2020
From: IKAI, TOMOHIRO; YAMAMOTO, TOMOYUKI; ITOH, NORIO; TOKUMO, YASUAKI
To: SHARP KABUSHIKI KAISHA
Reel/Frame 051725/0653 →
Priority Claims (2)
JP JP2017-155903 · Aug 10, 2017 · national
JP JP2018-053226 · Mar 20, 2018 · national
Continuity (1)
Related Publication 20210150767A1 · May 20, 2021
Cited By (1)
US 12,603,998