September 22, 2020

Google will launch a new image compression format, blessed by artificial intelligence

2 min read

Previously, Google developed a technology to use artificial intelligence for image compression. The technology is to let the artificial intelligence to 6 million randomly compressed 1280 × 720 pictures, and each divided into a number of 32 × 32 pixel blocks. After that, artificial intelligence in which to select the 100 compression rate of the most undesirable small pieces to learn.

After a lot of learning, the neural network through the complex operation can deconstruct the image, the image is divided into several small pieces, and for each piece of the actual situation “the right medicine” to take a tailored compression program, rather than using a unified method to compress the entire image.

This is the way Google used before. The use of artificial intelligence compression picture, can effectively filter out the unwanted parts, and truly optimize the effect.

Today, Google is planning to launch a new image format for small image services. In the future, this will likely become a new form of replacement for WebP.

According to 9to5Google reports, Google has Pik related documents uploaded to GitHub. This format is similar to the image format WebP published by Google in September 2010, which is the standard for compromised images of the network. 9to5Google said that the two properties are similar, Google is likely to use this new format to replace the previously released WebP.

However, the details of Pik have not been made public. It is not clear that it and web page specific differences, improvements are there. Pik’s documentation states that “this is not Google’s official product”, so Pik’s attribution is not determined.


At present, Pik is still in the unfinished state, so it is not certain what. But it can appear on Google’s official GitHub, which may also have some information. Or to wait for Pik development is complete, we will know the answer.