News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Artificial intelligence learning for low-pixel picture becomes clear
Artificial intelligence learning for low-pixel pic
We know the past two years, artificial intelligence technology to become the next outlet, many well-known technology companies competing for the high ground, the current Google's artificial intelligence learning technology let AlphaGo shine in the field of Go, according to foreign science and technology website arstechnica February 7 report , And now Google has developed a new software, to rely on artificial intelligence learning technology will become clear fuzzy picture.
From the examples given in the picture, Google this software to create the basis of the original image pixels, the original only 8 * 8 pixel size of the picture to restore the amazing details. We know that even if the artificial intelligence is not strong enough to create the original figure does not have the details, then the artificial intelligence technology in the end is how the photos become clear?
It is reported that to achieve this effect, the need for two neural networks together, the first phase involves the "conditioning network" (Conditioning Network), the system will search for other similar compressed high-resolution images, looking for patterns , Color similarity relationship.
The second phase, known as Prior Network, will use the details of high-resolution images to fill the pixels of the blurred image.
It is reported that the two neural network phase of the birth of the picture, will be further integrated into a picture, which contains a reasonable added details, in order to achieve the maximum high-resolution reduction.
In a test where 10 percent of the human observers were asked to look at a real celebrity high-resolution photo and a computer to restore the photo, Google's technology proved to be quite successful in the real world, And for the bedroom photos is up to 28% of human observers are confused by computer pictures.
It is worth noting that the computer super-resolution images are not real original images, the additional details in the field of image processing terminology is called "illusion", but the computer "guess", in other words the system that the new picture Clear enough to "deceive" the human eye, although not 100 percent reduction of the real, but in some areas such as the search for criminal suspects can still play no small supporting role.
In this case,
From the examples given in the picture, Google this software to create the basis of the original image pixels, the original only 8 * 8 pixel size of the picture to restore the amazing details. We know that even if the artificial intelligence is not strong enough to create the original figure does not have the details, then the artificial intelligence technology in the end is how the photos become clear?
It is reported that to achieve this effect, the need for two neural networks together, the first phase involves the "conditioning network" (Conditioning Network), the system will search for other similar compressed high-resolution images, looking for patterns , Color similarity relationship.
The second phase, known as Prior Network, will use the details of high-resolution images to fill the pixels of the blurred image.
It is reported that the two neural network phase of the birth of the picture, will be further integrated into a picture, which contains a reasonable added details, in order to achieve the maximum high-resolution reduction.
In a test where 10 percent of the human observers were asked to look at a real celebrity high-resolution photo and a computer to restore the photo, Google's technology proved to be quite successful in the real world, And for the bedroom photos is up to 28% of human observers are confused by computer pictures.
It is worth noting that the computer super-resolution images are not real original images, the additional details in the field of image processing terminology is called "illusion", but the computer "guess", in other words the system that the new picture Clear enough to "deceive" the human eye, although not 100 percent reduction of the real, but in some areas such as the search for criminal suspects can still play no small supporting role.
In this case,
PREVIOUS：Card business card recognition
NEXT：ID card OCR identification