We humans are visual creatures, and while video is the primary choice for online information consumption today, accounting for more than 80% of all online traffic, images have and will continue to play an important role in the digital experience. Whether we like to think about it or not, all the data bandwidth we consume has a cost, and industry giants like Google have a vested interest in optimizing said costs. Well, for better or worse, the search giant’s last-ditch effort in the image format space – WebP – didn’t go well and failed to achieve its goal of replacing JPG, PNG and GIF.
So Google changed its approach and decided to make JPEG better instead of fighting its ubiquity. Enter jpegli. It is a JPEG encoding library, including encoders and decoders. Perhaps the most important point is that both the encoder and decoder are compliant with “the original JPEG standard and its most traditional 8-bit form.” In simple terms, this means that images encoded using jpegli are compatible with existing codecs such as your browser or image viewer of choice.
Now, we won’t pretend that we know exactly what kind of “black magic” Google uses in jpegli. According to the press release, it “uses adaptive quantization to reduce noise and improve image quality. This is accomplished by spatially modulating the dead zone in quantization based on psychovisual modeling.”
However, some of the benefits outlined by jpegli are easier to grasp. For example, its image compression rate is approximately 35% higher than traditional JPEG codecs, while preserving the visual quality of the image. This alone is a huge win for online bandwidth. If nothing else, think about how much space Google itself could save in Google Photos by re-encoding user content. Additionally, jpegli apparently performs “more precise and psychovisually efficient calculations” so that images “look sharper and have fewer observable artifacts.” It can also encode images at 10+ bits per component compared to 8 bits of traditional JPEG encoding solutions, which occurs in the native 8-bit form without breaking compatibility with traditional 8-bit viewers, while Reduce “visible banding artifacts in slow gradients”. Last but not least, jpegli’s speed is obviously on par with other encoding libraries, so it doesn’t cost any extra computing resources or display processes.
A higher ELO score indicates better overall performance
Google has released the full jpegli source code on GitHub, so anyone interested can check it out and potentially start using it right away.
Source | Image source