In the latest issue of HackSpace magazine, out now, Ben Everard puts Raspberry Pi Camera Module 3 to the test.
The headline feature of almost any digital camera is the number of pixels, and all versions of the Raspberry Pi Camera Module 3 have 12-megapixel sensors – a 50% increase from the 8MP sensor on the Camera Module 2.
While most cameras are designed to take pictures for humans to look at, that’s not always the case with Raspberry Pi cameras. They’re just as likely to take photos that are analysed by AI, and there are different things that computers and humans take from images. For example, our jelly-filled-skull-holes might appreciate the crispness of an extra 50% of pixels, but to the cold silicon of an artificial intelligence, that means 50% more information to process. Feeding the full 12 million data points would choke all but the most powerful machine learning brains, but it does mean you can zoom into smaller areas than you could with the previous cameras. In this case, more pixels means more range.
While the extra pixels are a nice bonus, they’re – for us at least – not the most exciting things about the Camera Module V3. The two new features that we’re most interested in are the high dynamic range (HDR) support and autofocus.
When taking a photo, you have to expose the sensor for a certain amount of time. This, combined with the aperture (size of the opening that light passes through), determines how bright or dark the image is. The sensor has particular limits, so that if not enough light hits a pixel, it’s completely black. If too much hits it, it’s completely white. This causes a problem if part of the scene you want to capture is light and the other part is dark. If you set the exposure correctly for the light part, the dark part ends up being black (or very dark). If you set the exposure correctly for the dark part, the opposite happens.
HDR lets you apply different exposure settings to different parts of the image. Once upon a time, photographers would take multiple images, each with different settings, and then manually combine them to create one image. However, modern hardware – including the new Raspberry Pi Camera – can do this automatically. This mode is limited to 3 megapixels, but the results are typically much better-looking images than 12MP non-HDR images, especially in environments that aren’t specifically lit for photography. In our experiments, for anything other than cases where you needed the full resolution (such as print-ready images), HDR gives an easy visual boost to most images.
The new Camera Module brings in autofocus, which means that you can now capture images of things as close as 10 cm, or as far away as you like, without any problems.
Obviously, if only part of the image is in focus, this begs the question – what part? By default, the camera will attempt to keep the centre of the image in focus, but you can adjust this in your code. Whichever area you select, the camera will use phase detection to find the optimum lens position to keep it crisp and in focus. While shooting video, it can adjust as the subjects in the frame move.
One of the problems of reviewing the Raspberry Pi Camera Modules is the sheer range of uses that they can be used for. Some people use them to build cameras. Some people use them for AI robots. Some people use them as webcams. To help keep the camera as relevant as possible to as many people as possible, it comes in a few options. All of them have HDR and autofocus, but they have different lenses and filters.
The Camera Module V3 comes in normal and NoIR formats – the latter having the infrared filter removed, which means it captures more light and can be used with ‘invisible’ IR illumination to create a night-vision-like effect.
Both normal and NoIR have regular and wide-angle lens options. Wide-angle means that the camera captures a wider arc of the world in its image (120 degrees, rather than the standard 75). The flip side of this is that more of the world is spread across the available pixels.
Things like security cameras, and possibly robot cameras, will benefit from the wider field of vision, whereas, if you’re more interested in the things right in front of you, the standard lens will work better.
All these are the same size and connect in the same way, and they can all be controlled from software. However, the results are quite different, and which one is best for you depends a lot on what you want to do with it. Together, they offer a fantastic range of possibilities for embedded computing projects.
The Camera Module V3 is a huge improvement over the V2 in just about all areas of image quality, while still retaining the programmable control that makes it such a useful platform.
HDR and autofocus bring big image quality improvements to this maker camera.
Price: From $25
HackSpace magazine issue 64 out NOW!
As always, every issue is free to download in PDF format from the HackSpace magazine website.