A digital camera’s performance depends on many factors—the image sensor, the lens you use, and so on. Of these, one of the most important is the image processing engine, which is involved in almost all functions and processes during photography. On Canon EOS cameras, the core behind imaging excellence is the DIGIC image processing engine, developed by Canon in-house. Read on to learn more about it.
What is an image processing engine?
What is an image processing engine?
Before we can understand what an image processing engine does, let’s look at how a digital camera creates images.
How a digital image is produced: the simplified explanation
(1) Light from the scene enters the camera via the lens.
(2) The light is gathered by the image sensor.
(3) The image sensor encodes the information in the light into electrical signals.
(4) These electrical signals are processed by the image processing engine to form the digital image. In Canon EOS cameras, this image processing engine is called DIGIC.
As the diagram above shows, the main job of a digital camera's image processing engine is to help the image sensor translate light into a digital image. But it does more than just that. Digital cameras are made up of different components, such as the shutter unit and the parts that communicate with the lens, which all work together because they take instructions from the camera’s “brain”—the DIGIC image processing engine.
The history of DIGIC
The DIGIC image processing engine originated when Canon broke away from the norm of using mass-produced LSI (large-scale integrated) processors and decided to develop their own. The groundbreaking “imaging engine” that resulted was capable of processing much more data at high speed, and paved the way for more advanced, higher resolution cameras.
The first camera to be equipped with this imaging engine was the PowerShot S10, which was released in 1999. After multiple rounds of refinement, it was finally incorporated into the EOS 10D (released in 2003) as “DIGIC”.
What happens inside the camera when you shoot?
Have you thought about what happens inside the camera as you take a photo?
DIGIC, your camera’s multitasking brain
In reality, the image processing engine does a lot more than translate the light captured by the image sensor into a digital image.
For example, when we half-press the shutter button using an auto-exposure mode with AF (autofocusing) enabled, there are already at least two processes going on:
At shutter button half-press, the image processing engine starts metering the light. From there, it calculates the best exposure settings for the scene, and then communicates with the relevant parts to adjust the settings accordingly.
When you start the AF, DIGIC analyses the information from the image sensor to detect and track the subject. At the same time, it communicates with the lens to move the focusing mechanism inside the lens and establish focus.
Of course, that’s not the end of what DIGIC does. In fact, it’s just the beginning. The flowchart below shows how DIGIC is involved in every stage of image making, before and after you release the shutter:
Yes, that’s the amount of processing that is necessary for just one image!
Now imagine—if you shoot 10 frames per second, the above process will have to be repeated 10 times in one second. The higher the image sensor resolution, the more data to process per shot and the heavier the burden on the processor. That’s why a fast, powerful image processing engine is crucial for better camera performance.
How DIGIC has improved over the years
| 2002 | DIGIC | Achieved signal processing with just one processor chip |
| 2004 | DIGIC II | Faster, less noise |
| 2006 | DIGIC III | Higher resolution images; supports face detection |
| 2008 | DIGIC 4 | Higher resolution images; faster; detects movement |
| 2011 | DIGIC 5 | Better noise reduction; better white balance |
| 2013 | DIGIC 6 | Video capability (Full HD/60p) |
| 2016 | DIGIC 7 | Higher resolution images; more functionality (subject tracking, detection, image stabilisation |
| 2018 | DIGIC 8 | Improved shooting functionality and video capability (4K) |
| 2020 | DIGIC X | Improved video capability (>4K); faster; supports deep learning algorithm |
DIGIC and deep learning technology
DIGIC X: Supporting deep learning technology
The latest version of DIGIC is DIGIC X, which supports deep learning technology along with various other improvements.
Why is DIGIC necessary for deep learning technology?
Deep learning is a form of machine learning that makes use of many layers of artificial neural networks, which are based on similar structures in the human brain. The use of deep learning technology speeds up the development of subject recognition algorithms so that they can detect a wider scope of subject types with better accuracy.
EOS R7 + RF600mm f/4L IS USM @ f/4, 1/1600 sec, ISO 100
Deep learning is what helped train the algorithm in your DIGIC X-equipped camera to recognise a bird as a bird.
The cameras cannot “learn” on their own: this requires more processing power than can fit into a camera body. Instead, they are loaded with the deep learning algorithms that result from learning done at the development lab. Running these requires specialised circuitry and processing power, which DIGIC X has.
Cameras equipped with DIGIC X and what they can do
EOS-1D X Mark III: Debut of DIGIC X
The first camera to feature DIGIC X, the EOS-1D X Mark III, was capable of head detection via deep learning. This complemented existing face detection capabilities, improving subject tracking during scenes such as sports where an athlete’s face might become obstructed.
EOS R5 and EOS R6: Animal detection through deep learning
The EOS R5 and EOS R6 harnessed deep learning technology to detect the eyes, face, and whole body of cats, dogs, and birds.
EOS R3, EOS R7, EOS R10 and beyond
The EOS R3 used deep learning to support:
- Motorsports vehicle detection
- Human torso detection
- Head detection of winter sports athletes wearing goggles and helmets
- Enhanced eye detection
The EOS R3, and the EOS R7 and EOS R10 that followed it, also used deep learning technology to improve the accuracy of their auto white balance.
Learn more about the AF functions of the EOS R3 in:
Unravelling the AF Features on the EOS R3
What the cameras with DIGIC X can detect
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4-wheel drives/2-wheel drives |
|
|
|
|
|
|
●Video: Eye Detection AF on the EOS R3
DIGIC’s role during high-speed continuous shooting
High-speed continuous shooting and precise subject tracking
What does DIGIC do during continuous shooting?
Remember the processing flow outlined earlier in this article? During continuous shooting, that flow is repeated for every single shot taken. Additionally, DIGIC also performs calculations to predict the position of the moving subjects and keep them in focus. On the cameras that support deep learning-based AF, this includes running the processing-heavy deep learning algorithm.
The EOS R3 currently has the fastest continuous shooting capabilities of all EOS cameras: full resolution images at up to 30 fps with the electronic shutter when it was first released, and up to 195 fps (for up to 50 shots) with firmware update 1.2.0, announced in July 2022. Imagine: the entire processing workflow conducted 50 times in less than one second. This is testament to the speed and power of DIGIC X.
|
|
(max. burst speed) |
(max. burst speed) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* Up to 50 shots. With firmware update 1.2.0.
DIGIC, colour science, and image quality
The cornerstone of image quality
What does the image processing engine do after you take a photo?
Let’s recap the process described in the beginning of the article. In reality, Part (4) of the process is a lot more complex. To put it into layman’s terms, it looks more like this:
The RAW data is like the digital version of film negatives—it needs to be digitally developed (processed) before we can view the image. It is the DIGIC image processing engine that controls the image sensor, reads the electrical signals, and does the digital image development.
Know this: DIGIC’s role in colour science
On its own, the image sensor is “colour blind”—it can only capture information about the strength of the light reaching it, but not the colour. Colour images are achieved mainly due to:
1. An RGB colour filter array (CFA) located in front of the image sensor, which filters the light into separate red (R), green (G), and blue (B) components. The image sensor can’t differentiate the colours on its own, but it can capture the strength of the light in each component, which it then transmits to the image processor as electrical signals. These signals are also encoded in the RAW file as digital data.
2. A debayering (or demosaicing) algorithm, which the image processing engine uses to process the R, G, and B data from the image sensor, and then render (“reconstruct”) the colours in the image.
These are the two main factors behind what many refer to as a camera’s “colour science”. How well colours are rendered straight out of camera reflects how well an image processing engine handles colour data.
What happens when you remove the colour filter from the image sensor? Find out here—but don’t try this on your own!
RAW development and processing
When DIGIC digitally develops the RAW data, it generally conducts 14-bit processing, including demosaicing the RGB data to “reconstruct” the colours. At the same time, it also conducts noise reduction, and adjusts sharpness, contrast, colour tones, and white balance.
The EOS R3 achieves native ISO speeds of up to 102,400 because of a new noise reduction processing algorithm in DIGIC X.
DIGIC also conducts post-processing to improve the quality and visual appearance of the straight out of the camera image. These include applying:
- Lens aberration correction
- Creative filters
- In-camera HDR merging
- Multiple exposures
Image processing is complete when the image is converted to JPEG (8-bit) and/or HEIF (10-bit), compressed, and written to the memory card. Now you know why your straight-out-of-camera JPEG or HEIF files look so much better than RAW files!
HDR PQ HEIF: HDR images in just one exposure
Creating a realistic HDR (High Dynamic Range) image on a single 8-bit JPEG file usually requires merging multiple exposure-bracketed shots. However, on EOS cameras that use DIGIC X, you can do in just one exposure by recording in 10-bit HDR PQ HEIF.
Find out more in:
HDR PQ HEIF: Breaking Through the Limits of JPEG
The processing power that enables 8K video
Nowadays, high-resolution video recording has become an essential function on all digital cameras. However, it takes tremendous processing power to capture high-quality, high-resolution video due to the amount of data in each frame.
Until DIGIC 7, Canon had mainly prioritised still photography when developing its image processing engines. However, from DIGIC 8 onwards, it began to focus more on video-related processing functions. This gave rise to cameras with better video capabilities than before, from 4K shooting to support for more video features.
Key video specifications for cameras with DIGIC X
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* With firmware update 1.2.0.
One 4K DCI/ UHD frame is 8.85 and 8.3 megapixels respectively. What that is much less than one still photo on the same camera, AF and autoexposure must be recalculated for every single frame. This means that the processing flow introduced above is repeated 60 times per second when you shoot a 60 fps video.
Lest you think it’s because 4K is small, almost all the DIGIC X cameras can shoot oversampled 4K—which means they are handling more data than 4K. In fact, an 8K frame is four times larger than a 4K frame, and the EOS R5 C is capable of 8K DCI 60P RAW recording. This reflects the capabilities of DIGIC X.
Also see:
EOS C70, R5 C, R5, or R3: Which to Get for Video?
What are 8K, 4K, and Full HD? How Do I Use Them?
In summary: DIGIC, the core of camera performance
When discussing a camera’s performance, many people pay attention to features such as the image sensor’s dynamic range. However, we shouldn’t neglect the image processing engine, whose performance determines the final image quality.
In Canon’s pursuit of high image quality, DIGIC has undergone years of refinement. Its latest version, DIGIC X, was developed with both still photography and videography in mind. With its specialised circuitry for processing deep learning algorithms as well as enhancements that speed up its various processes, it is a core component vital for achieving outstanding image quality and a seamless shooting experience.
.jpg "trending tips")
.jpg "trending tips")
.jpg "trending tips")
.jpg "trending tips")
.JPG "trending tips")
.JPG "trending tips")
