A Brief Analysis of Starlight-Level Low-Illumination Technology

    Although traditional infrared fill-in technology can obtain clear images under low illumination, it will lose color and can only form black and white images. In addition, through infrared fill-in, highly reflective objects such as license plates are easily overexposed, and clothing color, body color, license plates, etc. are often key clues to solving a case and cannot be lost - so the demand for clear color imaging under low illumination is increasing.
       
    With the introduction of Axis' "Lightfinder" technology, starlight-level low-illumination imaging technology has begun to develop and gradually mature. More and more companies have launched various models of high-definition starlight-level cameras. For example, the corresponding starlight-level low-illumination models have been added to the HIC5400 series and HIC5600 series IPCs of my company. It can be said that the past 2014 was a year of great development of starlight-level low-illumination technology. So what is its effect?

　　

 Ordinary IPC low light effect

Star-level IPC low-light effect

　　 As shown in the comparison above, this is a comparison of the effects of an ordinary camera and a Uniview starlight-level low-light camera at the same time at night in a park where the street lights are not turned on. The image obtained by an ordinary camera after forced color can only vaguely see the outlines of trees, roads, and buildings. The overall environment is dark and has lost its monitoring function; while Uniview's starlight-level low-light camera can restore the details of roads, buildings, and even distant vehicles very clearly, and there is no obvious noise in the overall picture, and the color effect under low light is excellent.
         
    In fact, the structure of the starlight-level low-light camera is exactly the same as that of an ordinary camera. The difference in its effect mainly comes from the improvement of the underlying hardware performance and the further optimization of the image adjustment algorithm.
         
    First of all, the sensor is the factor that has the greatest impact on the low-light effect of IPC. At present, mainstream IPCs all use CMOS sensors. Although at the beginning, due to the differences in the structures and working mechanisms of the two sensors, the imaging effect of CMOS under low light is not as good as that of CCD, but with the introduction of back-illuminated CMOS, its internal structure has been adjusted, and the light can directly reach the photosensitive layer after passing through the lens, which makes the sensitivity have a qualitative leap. The sensor sensitivity is the biggest guarantee for the camera to still obtain clear color images in weak ambient light.
       
    Secondly, the low-light effect of the camera is also affected by the lens. The maximum aperture value of the lens is a very important factor. The larger the aperture, the more light enters, and the better the color effect. On the contrary, the effect is worse. However, the larger the lens aperture, the shallower the depth of field, that is, the smaller the focus range in the scene. Therefore, the lens aperture value also needs to be adjusted according to the actual scene. The aperture cannot be increased blindly for the amount of light entering. Another similar parameter is the camera slow shutter. When the shutter is slower, the amount of light entering is greater, the brightness of the picture is improved, but the relative smear of moving objects is also more serious. Therefore, the shutter value is also a double-edged sword, and the corresponding value needs to be set according to the specific application scenario.
         
    Finally, the low-light effect is also strongly related to the processing effect of the ISP. For example, noise reduction processing is a very important processing process for imaging under low illumination. Noise will make the image appear unclear, but improper noise reduction processing will cause serious fog, smear, loss of details and other phenomena. Therefore, even if based on the same hardware platform, the final effect of the product is different, which reflects the R&D capabilities of different manufacturers.
         
    At present, in actual projects, the most common application scenario of low-light cameras is on the road, where it is necessary to see the overall environment clearly, suppress the headlights of passing vehicles, and see the license plate clearly. This is a new challenge faced by starlight cameras, because on the one hand, in low-light scenes, as much light as possible needs to be allowed into the camera, but on the other hand, the overly bright light needs to be weakened, otherwise it is easy to overexpose and interfere with the imaging effect. However, high-end starlight cameras have solved this problem. The following figure shows the image effect of Uniview's starlight gun camera in a project. The strong light of the car headlights is suppressed, and the details of the car body including the license plate are clearly visible, and the surrounding road environment is also very clear.

　　

Star-level IPC strong light suppression effect

　　Another problem is that as the number of pixels increases, the color low-light effect becomes worse. This is because the camera has requirements for the size of the sensor. When the size does not change much, the more pixels there are, the smaller the photosensitive area per pixel, and the worse the low-light effect. Currently, starlight cameras are mainly 1080P resolution, all based on sensors of about 1/2", and the mainstream 300W cameras on the market are based on 1/3" sensors. Naturally, it is difficult to obtain good low-light effects at night. For higher-pixel cameras such as 4K cameras, the sensor size is only about 1/2", so the low-light effect is naturally not good. However, with the increasing calls for 4K ultra-high-definition IPCs and more and more applications, the combination of high resolution and starlight technology is bound to be the general trend. I believe this will be an important direction for the development of starlight technology.