Clearer Pictures of Black Holes Produced Using Machine Learning

 

The first picture of a black hole taken in 2019 confirms several theories about what it looks like for decades. At that time the resulting image looked blurry but clearly showed the shape of an orange donut with a black hole in the middle. Images taken by the Event Horizon Telescope (EHT) have been reprocessed using new machine learning algorithms to produce clearer images.

The same image was processed by the principle-component interferometric modeling (PRIMO) algorithm which was trained using 30,000 black hole simulation images produced by different models. The result is a noise-reduced image. The orange ring is also smaller with the black hole in the center more visible.

This clearer image will help scientists understand more about black holes. It can be used to see how space and time warp when approaching high gravitational pull.