Method of evaluating crosstalk and crosstalk evaluation device performing the same
A method of evaluating crosstalk includes receiving a first multi-view 3-dimensional image, a second multi-view 3-dimensional image, and a third multi-view 3-dimensional image, performing a deep learning on a crosstalk evaluation for classifying the first multi-view 3-dimensional image into a normal image or a defective image to generate a learning model, and inputting the third multi-view 3-dimensional image different from the first multi-view 3-dimensional image into the learning model to perform the crosstalk evaluation of the third multi-view 3-dimensional image.
1. A method of evaluating crosstalk, the method comprising:
providing, by a display panel, a first multi-view 3-dimensional image, a second multi-view 3-dimensional image, and a third multi-view 3-dimensional image to a diffuser screen;
acquiring the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image reflected by the diffuser screen;
receiving the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image;
performing a deep learning on a crosstalk evaluation for classifying the first multi-view 3-dimensional image into a normal image or a defective image to generate a learning model; and
inputting the third multi-view 3-dimensional image different from the first multi-view 3-dimensional image into the learning model to perform the crosstalk evaluation of the third multi-view 3-dimensional image.
2. The method of claim 1 , further comprising: inputting the second multi-view 3-dimensional image different from the first multi-view 3-dimensional image and the third multi-view 3-dimensional image into the learning model to verify the learning model.
3. The method of claim 1 , wherein the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image provided by the display panel are controlled by an optical device, and the optical device is a lenticular lens or a parallax barrier.
4. The method of claim 3 , wherein the defective image is generated by misalignment between the display panel and the optical device.
5. The method of claim 1 , wherein the defective image includes at least one of a low brightness image, a background noise image, a spreading image, and a curved image.
6. The method of claim 3 , further comprising: realigning the display panel and the optical device in case that the defective image is generated by misalignment between the display panel and the optical device.
7. The method of claim 1 , further comprising: compressing the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image such that resolutions of the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and third multi-view 3-dimensional image are resized from a first resolution to a second resolution lower than the first resolution.
8. A method of evaluating crosstalk, the method comprising:
providing, by a display panel, a first multi-view 3-dimensional image, a second multi-view 3-dimensional image, and a third multi-view 3-dimensional image to a diffuser screen;
acquiring the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image reflected by the diffuser screen;
receiving the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image;
extracting a grayscale of the first multi-view 3-dimensional image;
performing a deep learning on a crosstalk evaluation for classifying the first multi-view 3-dimensional image into a normal image or a defective image to generate a learning model; and
inputting the third multi-view 3-dimensional image different from the first multi-view 3-dimensional image into the learning model to perform the crosstalk evaluation of the third multi-view 3-dimensional image.
9. The method of claim 8 , further comprising: inputting the second multi-view 3-dimensional image different from the first multi-view 3-dimensional image and the third multi-view 3-dimensional image into the learning model to verify the learning model.
10. The method of claim 8 , wherein the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image provided by the display panel are controlled by an optical device, and the optical device is a lenticular lens or a parallax barrier.
11. The method of claim 10 , wherein the defective image is generated by misalignment between the display panel and the optical device.
12. The method of claim 8 , wherein the defective image includes at least one of a low brightness image, a background noise image, a spreading image, and a curved image.
13. The method of claim 10 , further comprising: realigning the display panel and the optical device in case that the defective image is generated by misalignment between the display panel and the optical device.
14. The method of claim 8 , further comprising: compressing the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image such that resolutions of the first multi-view 3-dimensional image, the second multi-view 3-dimensional image and the third multi-view 3-dimensional image are resized from a first resolution to a second resolution lower than the first resolution.
15. A crosstalk evaluation device comprising:
a display panel;
a diffuser screen that reflects a first multi-view 3-dimensional image, a second multi-view 3-dimensional image, and a third multi-view 3-dimensional image provided by the display panel;
a digital camera that acquires the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image reflected by the diffuser screen; and
a crosstalk evaluator that receives the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image, performs a deep learning on a crosstalk evaluation for classifying the first multi-view 3-dimensional image into a normal image or a defective image to generate a learning model, and inputs the third multi-view 3-dimensional image different from the first multi-view 3-dimensional image into the learning model to perform the crosstalk evaluation of the third multi-view 3-dimensional image.
16. The crosstalk evaluation device of claim 15 , wherein the first multi-view 3-dimensional image, the second multi-view 3-dimensional image, and the third multi-view 3-dimensional image provided by the display panel are controlled by an optical device, and the optical device is a lenticular lens or a parallax barrier.
17. The crosstalk evaluation device of claim 16 , wherein the defective image is generated by misalignment between the display panel and the optical device.
18. The crosstalk evaluation device of claim 15 , wherein the defective image includes at least one of a low brightness image, a background noise image, a spreading image, and a curved image.