REAL-TIME MASK QUALITY PREDICTOR

• US 20190325578A1
• Filed: 06/28/2019
• Published: 10/24/2019
• Est. Priority Date: 06/28/2019
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Mohammad, Fahim, Batz, Joseph, Segerlind, Nathan, Benou, Itay, Hershkovich, Tzachi
• Granted Patent: US 11,164,317 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 7/10 Segmentation; Edge detection motion-based segmentation G06T7/215
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation

Abstract

An embodiment of mask quality predication technology may include a memory to store a set of input images and a reference mask image associated with each input image of the set of input images, a processor communicatively coupled to the memory, and logic communicatively coupled to the processor and the memory to generate a set of two or more masks of different quality associated with each input image of the set of input images, and determine a quality score for each generated mask. Other embodiments are disclosed and claimed.

Claims

1. An image processing system, comprising:

a video capture engine to capture a volumetric broadcast video signal in real-time and generate a sequence of frame images from the captured real-time volumetric broadcast video signal;

an image segmentation engine to segment an input image, which corresponds to a single frame of the sequence of frame images, to generate a mask image associated with the input image; and

logic coupled to the image segmentation engine and the video capture engine to determine a mask quality score based on the input image and the associated mask image in real-time.

2. The system of claim 1, wherein the logic comprises:

a neural network trained to take the input image and associated mask image as inputs to the neural network and provide the mask quality score as an output of the neural network.

3. The system of claim 1, wherein the neural network is trained based on a set of reference input images, a set of degraded mask images, and a set of quality scores.

4. An electronic system, comprising:

a memory to store a set of input images and a reference mask image associated with each input image of the set of input images;

a processor communicatively coupled to the memory; and

logic communicatively coupled to the processor and the memory, the logic to;

generate a set of two or more masks of different quality associated with each input image of the set of input images, and

determine a quality score for each generated mask.

5. The system of claim 4, wherein the logic is further to:

generate at least one mask of the set of masks with a first segmentation model; and

generate at least one mask of the set of masks with a second segmentation model which is different from the first segmentation model.

6. The system of claim 4, wherein the logic is further to:

generate at least one mask of the set of masks with a first set of parameters for a segmentation model; and

generate at least one mask of the set of masks with a second set of parameters for the segmentation model which is different from the first set of parameters. The system of claim 4, wherein the logic is further to;

determine the quality score for each generated mask based on a comparison of the generated mask associated with an input image and the reference mask associated with the input image.

8. The system of claim 7, wherein the comparison is based on a Dice similarity coefficient.

9. The system of claim 7, wherein the logic is further to:

train a neural network to predict quality scores based on a training data set that includes the set of input images, the set of two or more generated masks associated with each input image, and the quality score for each of the generated masks.

10. An image processing apparatus, comprising:

one or more substrates; and

logic coupled to the one or more substrates, the logic to;

generate a set of two or more masks of different quality associated each input image of a set of input images, and

determine a quality score for each generated mask.

11. The apparatus of claim 10, wherein the logic is further to:

generate at least one mask of the set of masks with a first segmentation model; and

generate at least one mask of the set of masks with a second segmentation model which is different from the first segmentation model.

12. The apparatus of claim 10, wherein the logic is further to:

generate at least one mask of the set of masks with a first set of parameters for a segmentation model; and

generate at least one mask of the set of masks with a second set of parameters for the segmentation model which is different from the first set of parameters.

13. The apparatus of claim 10, wherein the logic is further to:

determine the quality score for each generated mask based on a comparison of the generated mask associated with an input image and a reference mask associated with the input image.

14. The apparatus of claim 13, wherein the comparison is based on a Dice similarity coefficient.

15. The apparatus of claim 13, wherein the logic is further to:

train a neural network to predict quality scores based on a training data set that includes the set of input images, the set of two or more generated masks associated with each input image, and the quality score for each of the generated masks.

16. A method of processing an image, comprising:

generating a set of two or more masks of different quality associated each input image of a set of input images; and

determining a quality score for each generated mask.

17. The method of claim 16, further comprising:

generating at least one mask of the set of masks with a first segmentation model; and

generating at least one mask of the set of masks with a second segmentation model which is different from the first segmentation model.

18. The method of claim 16, further comprising:

generating at least one mask of the set of masks with a first set of parameters for a segmentation model; and

generating at least one mask of the set of masks with a second set of parameters for the segmentation model which is different from the first set of parameters.

19. The method of claim 16, further comprising:

determining the quality score for each generated mask based on a comparison of the generated mask associated with an input image and a reference mask associated with the input image.

20. The method of claim 19, wherein the comparison is based on a Dice similarity coefficient.

21. The method of claim 19, further comprising:

training a neural network to predict quality scores based on a training data set that includes the set of input images, the set of two or more generated masks associated with each input image, and the quality score for each of the generated masks.