Mosaic is a widely used image (video) processing method, which worsens the color scale details of specific areas of the image and causes color blocks to be disturbed. Because this blur seems to be composed of small squares, it is vividly called mosaic. Its purpose is usually to make it unrecognizable.
At present, mosaics are mostly text pictures and ordinary pictures. In the picture, with the current technology, if it is only a small part, we can restore it to a certain extent through technical means, but it needs a lot of original paintings to restore, and generally no one does it.
Extended data
At present, the restoration of mosaic belongs to the category of artificial intelligence. For example, Google developers make AI "associate" through the AI system, add details to the mosaic through "guessing", gradually increase the pixel multiple, and finally get a clear image. This processing method requires AI's powerful learning ability and computing ability, which is also the development direction of artificial intelligence.
This "brain tonic" is based on two core tools. The first item is the tool adjustment network, which is a CNN (Convolutional Neural Network). It trains by showing a large number of facial photos to AI, so that AI can learn representative facial features through memory.
The second tool is the priority network, which is a pixel CNN. It allows the system to guess which details can be used as the characteristics of high-resolution photos according to previous training, and find the most likely matching details from all possible original images to fill high-resolution photos according to the principle of probability priority.
At present, the restoration technology can only fill, match and restore the contents of the original images in countless images through a huge database under the accurate computing power of AI, but for the following contents, no matter how powerful the technology is, it can't be restored without database information, so there is no technology to restore this text mosaic in the market at present.