Image quality in the field of industrial image processing

For senior photographers, good image quality means an image with sharpness, high brightness and high contrast. In the field of industrial image processing, image quality includes more dimensions.

Suppose you have two industrial cameras that meet your needs and look the same from the outside, how can you tell which camera can get better picture quality?

Let's first look at the data table of EMVA1288. EMVA1288 (web site: http://www.emva.org) is a standard developed by the European Commission on Machine Vision to evaluate the image quality and sensitivity of cameras and sensors.

1288 standard data sheet. png

Many of the values in the tables are good, but is the higher the value, the better? Many people may be familiar with sensors and pixel values, and feel that they are the most important factors affecting image quality.

Sensor and Pixel Size, Quantum Efficiency, Resolution

Sensor is the core of the camera, because it is really the most important component. Sensors contain many photodiodes that carry pixel information, which is converted from photon energy to electronic information, which is then converted into images that can be processed by software. There is a proportion of photon information obtained from electronic information, which is called Quantum Efficiency QE in%. Therefore, under the same amount of photon information, the more electronic signals are generated, the higher the quantum efficiency, and the higher the quantum efficiency, the better the image quality.

Take the SLR digital camera as an example, it has large sensors, while the ordinary camera has smaller sensors. Ideally, both cameras can produce high quality images. But when it comes to dark environment, the image quality of SLR camera is better. Large sensors can get more photon information, while ordinary cameras can quickly compensate for the information they don't get by flashing lights. If you set up an ordinary camera without flash, it will adjust the ISO, which will increase the image noise.

Simply put, the bigger the sensor, the higher the image quality. Here's a comparison of sensor sizes

Sensor Size Contrast.png

Then, besides the influence of sensor size on image quality, the number of pixels and the size of sensor surface are also important factors affecting image quality.

A single-mirror camera (full-size) and a conventional camera (4/3") have 15 million pixels. Obviously, if the sensor size is unchanged, the pixels in the sensor will be smaller in the ordinary camera in order to obtain the same resolution. If there are too many pixels on the surface of the sensor, it will cause more noise and less light sense, which will be too large. The quality of the image is also deteriorating.

In fact, pixels usually combine sensors, lenses and distances between objects to determine the size of the pixels is crucial to the amount of light information obtained.

For example, it's like millions of small barrels of iron followed by rain. The larger the capture area, the more photon information it collects. Through exposure, the more precise the values in each barrel are, we usually have 256 different gray values, 0 for all black, 255 for all white, just as the sensor perceives the values. Once the rainwater in the bucket is full, that is, the pixels have been collected, the overflow information will be processed in 255, that is, the image information is lost.

Shadow section. png

For example, in this picture, the shadows need a long exposure time to collect enough photon information, because the details are still not delicate enough, while the clouds are overexposed, like a bucket overflowing after filling. If you compress the exposure time now, the normal pixels will not have time to get enough photon information, and these pixels will become a little darker. This is why SLR cameras have better image quality.

In fact, the camera with large pixels can get more gray detail information.

dynamic range

We have an indicator of how much detail a camera's image can get, called dynamic range, which is a proportional range value to measure light sensitivity. In the example of the previous picture, the dynamic range of the camera can satisfy the need for light and dark details.

In the application of industrial vision system, the camera in traffic scene needs a wide dynamic range in order to obtain the vehicle speed.

In the area next to the license plate, it often appears brighter, while the area next to the driver appears darker.

A moving car.png

Cameras with wide dynamic range can get clear images of different areas (head and driver), while cameras with narrow dynamic range can only get clear images of a single area (head or driver).

SNR

Noise is another key indicator of image quality. Look at the enlarged particles in this picture. They have obvious noise.

We use the signal-to-noise ratio to measure image quality. Suppose we listen to music while driving. We think the music in the car is a signal. Besides, there are a lot of noise. We compare them to get the signal-to-noise ratio. The better the music effect is, the higher the signal-to-noise ratio is, the worse the effect is, and the lower the signal-to-noise ratio is.

Back to our examples of SLR cameras and ordinary cameras, SLR cameras have a better signal-to-noise ratio, which obtains more photon information and less noise information.

The higher the signal-to-noise ratio, the less the particle sensation in the image and the higher the image quality.

Sensor technology plays an important role here. CMOS sensor technology has developed well in recent years. It has the following advantages

Powerful

High sensitivity

fast

Low price

But only machine manufacturers have special notes.

Please read the Chinese version for details.