Alphabet Inc.’s latest project doesn’t have such limitations. ‘Guetzli’ is a new image encoder out from Google Research’s office in Zurich, translating to ‘cookie’ in English. The encoder offers image compression for smaller or better-looking images. Even better, Even better, it’s open source.Developers can view the Guetzli code on Github and integrate it into their own projects. If it takes off, we could see image processing apps ditching the current jpeglib encoder and moving to Google’s method.
How it Works
Though there are plenty of different methods for compression, Guetzli has a particular focus on quantization. It takes disordered data and orders it, making it much easier to compress. However, in most methods this results in a loss of quality, resulting in blocky images instead of smooth ones. As each image has different requirements, it’s hard to create an encoder with a minimal visual impact that also reduces file size sufficiently. As a result, Google’s research team created a new model called ‘Butteraugli’. According to the team, the Butteraugli model is based on the way humans perceive the world. It can apparently guess color perception and other factors in a more nuanced way: “Guetzli strikes a balance between minimal loss and file size by employing a search algorithm that tries to overcome the difference between the psychovisual modeling of JPEG’s format, and Guetzli’s psychovisual model, which approximates color perception and visual masking in a more thorough and detailed way than what is achievable by simpler color transforms and the discrete cosine transform,” said the team. Of course, such functionality comes at a price, and in this case, it takes “significantly longer” to compress images. However, for the websites of photographers and graphic designers, the payoff will likely be worth it. You can download Guetzli today from GitHub.
