Digital image sensors work based on the Photoelectric effect. When the image sensor is exposed to light (photons), an equivalent charge is produced by the image sensor. It is then converted to image data.
Vision in Challenging Lighting Environments
Contributed by | e-con Systems
e-con Systems recently launched its Ultra Low Light High Dynamic Range USB 2.0 camera Hyperyon.
.jpg)
The block flow diagram of Hyperyon is shown below.
The UVC compliant camera is based on IMX290 - a popular sensor from Sony's STARVIS range. It is a 1/2.8" 2 MP back-illuminated CMOS image sensor with excellent sensitivity in the visible-light and near-infrared light regions.
This camera sports a sophisticated ISP by Socionext, which delivers high-quality images with low noise even at lighting conditions below 1 Lux.
So, what makes e-con's Hyperyon the ideal camera for a challenging lighting environment? Let us look at some of its salient features.
High Sensitivity - Capture Images at Near Darkness
The image sensor used in Hyperyon has very high quantum efficiency, which makes it extremely sensitive to light and a perfect choice to boost your surveillance applications.
What is quantum efficiency?
Digital image sensors work based on the Photoelectric effect. When the image sensor is exposed to light (photons), an equivalent charge is produced by the image sensor. It is then converted to image data.
The ratio of an incident photon to a converted electron is known as quantum efficiency. Sensors with high quantum efficiency can capture detailed images at very low illumination when compared to sensors with low quantum efficiency. The low light efficiency of Hyperyon can be understood from the below images.
|
Other Cameras |
Hyperyon |
|
|
|
|
Captured at 0.50 Lux at 33 ms |
Captured at 0.50 Lux at 33 ms |
|
Captured at 0.05 Lux at 33 ms |
Captured at 0.05 Lux at 33 ms |
.jpg)
3D Noise Reduction
Noise is an unwanted signal. Image noise can appear as a random variation of brightness or color information, which degrades the image quality. There are many types of noises like Gaussian noise, shot noise, and salt and pepper noise. Each originates from a different source.
The effect of these noise on the image is based on factors such as gain, exposure time, temperature, and others. Noise is an unavoidable component in every electronic device which receives or transmits a signal. However, the effect of noise is minimal since the strength of the signal is higher when compared to the strength of the noise.
The Signal to Noise Ratio (SNR) is a universal way of comparing the relative amounts of signal and noise for any electronic system.
Why images captured in very low light tend to be noisy?
A camera adjusts the exposure time and gain of the sensor according to the lighting conditions. When the scene is well illuminated, the exposure time and gain are set low to expose the scene properly.
Similarly, in low light, the camera increases the exposure time of the sensor to capture more details. Since the signal strength is low in low light conditions (fewer photons to capture), the gain of the sensor is increased to amplify the signal.
As the lighting condition reduces, so does the strength of the image signal. Since there isn't enough light hitting on the sensor, the only available signal for the sensor to sense is noise.
The effect of noise worsens when the gain increases because the noise signal also gets amplified. In a well-illuminated environment, the noise signal is masked by the image signal because of ample light, and the effect of noise is very minimal.
How does Hyperyon help with noise reduction?
Hyperyon uses a sophisticated 3D noise reduction algorithm, which performs spatial and temporal noise reduction on the image to reduce the effect of noise.
Spatial noise reduction is the process of noise removal from each frame. Temporal noise reduction is a complex noise reduction process, which involves the removal of noise between frames over time, as shown below.
.jpg)
Hyperyon ensures that the captured images have maximum details and minimum noise which makes it an ideal choice for low light imaging, as shown below.
|
Other Cameras (Without Denoise) |
Hyperyon (With/Without Denoise) |
|
Captured at 33 ms Gain 66 dB |
Captured at 33 ms Gain 30 dB without denoise |
|
Captured at 250 ms Gain 66 dB |
Captured at 33 ms Gain 30 dB with 2DNR |
|
Captured at 500 ms Gain 66 dB |
Captured at 33 ms Gain 30 dB with 3DNR |
High Dynamic Range
The most common problem while capturing the image is proper exposure. The camera's auto exposure algorithm tries to expose the sensor to optimum exposure time to make sure the image is not overexposed or underexposed.
But when light and dark regions coexist, the camera's auto-exposure fails to exposure all the details in the frame correctly. Few parts of the captured image are either underexposed or overexposed.
Hyperyon uses HDR imaging technology, which enables both the light and dark regions to be properly exposed in the captured image, as shown below.
|
Other Cameras |
Hyperyon - Never miss a detail |
|
|
|
|
|
|
View HDR Comparison video:
https://www.youtube.com/watch?v=FSfSQjF_4GE
On-board Encoder - High on Quality, Low on Size
Hyperyon has an on-board H264 and MJPEG encoder, which gives it the power to deliver high-quality encoded frames at 60 fps.
What is the main advantage of encoding video data?
Imagine a surveillance application where video data must be saved for long hours. If the user wants to save the data without encoding, he/she will require a lot of storage space. The same case applies to video calling applications. Transmitting a non-encoded video over a network will consume a lot of bandwidth. To overcome this issue, the user must use an effective encoder at the host to reduce the size of the video stream with minimal reduction in quality, which might need extra computation - subject to availability.
Hyperyon has the option to control the quality of the image to be encoded. With this feature, the user can reduce the size of the image reasonably without compromising the quality of the image. The encoded video stream can be used for image processing algorithms that require high levels of detailing like object detection and tracking, face detection, and more.
Comparison of various encoder images
|
YUV |
MJPG |
H264 |
|
|
|
|
|
|
|
|
If you need any other information on Hyperon, please write to marketing@e-consystems.com.
The content & opinions in this article are the author’s and do not necessarily represent the views of RoboticsTomorrow
Comments (0)
This post does not have any comments. Be the first to leave a comment below.
Featured Product
