Table of contents:
Portrait mode has become, in its own right, one of the features that users most surprise and crave when they have to opt for one mobile phone or another. Photographs with this portrait mode, in which an object, person or animal appears in the foreground, while the background remains blurred. To achieve this effect, manufacturers choose to include a combo of two lenses in the phone so that each one does its job separately. However, Google wanted, for its range of Pixel phones, to test with a single camera and create that effect thanks to post-processing using artificial intelligence technology.
The Google Pixel 3 wants to achieve with one camera what others with three
However, this method to achieve portrait mode produced some flaws that would try to be corrected in the next Google Pixel 3 in a somewhat radical and, at least, curious way. To train its artificial intelligence technology, Google built a kind of mutant mobile, composed of five phones linked together, leaving the camera viewfinders unobstructed, in order to take different photographs from different perspectives with slight differences. These small differences then allow the computers that analyze them to determine how 'far' one photograph is from another, thus generating a 'depth map', used later to 'draw' the background that would then be extracted to blur it.
The Google team that built such a peculiar contraption saw fit to name it 'Frankenphone', thus alluding to the famous Doctor who gave life to 'the creature' through pieces of different corpses. Researcher Rahul Garg and programmer Neal Wadhwa spoke on the Google blog last Thursday:
It is a fact that mobile phone cameras cannot yet compete with the traditional imaging systems that we can see in professional cameras. Those still have small image sensorsthat do not reach the superior quality of a reflex equipment. Still, the distances are shortened thanks to these ingenious methods that combine software and hardware. Google wants to stay at the forefront in mobile photographic innovation thanks to 'computational photography' methods that achieve blurry backgrounds, increase image resolution, adjust exposure, improve shadow detail, and take photos in low light. Another thing is that it achieves it, taking into account that it faces examples such as the Samsung Galaxy Note 9 and its camera with double focal aperture or the Huawei P20 Pro and its triple photographic sensor.
With this 'Frankenphone', Google researchers want to transfer our vision of the world to the photographic section. Human beings have two eyes, located at a certain distance, that offer a deep vision of the world. This is what has been tried to transfer to the photographic sensor of the Google Pixel 2 and Google Pixel 3. Each pixel of a photograph taken by one of these two terminals is created by two light detectors, located on the left and right. This difference in distance between the two detectors emulates the distance of our eyes and thus manages to give a feeling of depth and create portrait mode without the use of two sensors at the same time.
