In the last few months, starting with the Mix 2s, Xiaomi has started to get its act together in the photography space, followed by the Mi 8, and then the Mix 3, and the company has proven that it can excel in this area too. It's not just the number of megapixels that determines the character of an image, but the digital processing, which is now being significantly improved with the introduction of AI. This is further enhanced by the newly introduced DeepExposure AI algorithm, which achieves even better results than before.
As we know, the camera on mobile phones has the problem of capturing poor quality images when there is too much or too little light. Apart from a few manufacturers, these problems are still present in all but the most expensive handsets today, which bypass the problem somewhat. Xiaomi has extended its AI with the DeepExposure AI algorithm to achieve even more perfect lighting and detail.
Xiaomi DeepExposure balances the exposure by dividing the photo into sub-registers and then adjusting the exposure value of each fragment separately. Finally, the different fragments are combined to form the final image. You can see the general process in the image below.
The new algorithm splits the whole process into image segmentation and "action generation". After segmentation, the second part starts with input resolution, sub images, and direct fusion of the image takes place. All photos are processed with each local and global exposure.
The new algorithm was created on the Nvidia P40 Tesla GPU in Google's TensorFlow framework. The Xiaomi team used images from the MIT-Adobe FiveK dataset, making Xiaomi's new DeepEposure AI algorithm outperform all other algorithms on the market. From this, it is hoped that this new algorithm will help Xiaomi devices to outperform the camera's previous performance and other methods.
