In nature, according to the human naked eye can feel the light band range, divided into visible light and invisible light. The wavelength range of visible light is 400nm~780nm, the wavelength range of invisible light is 10nm~400nm, 780nm~0.1mm.
01 Color of visible light
In nature, the white light (sunlight) perceived by the human eye is actually made up of many different colors of light, and each color of light has a different frequency.
Since human eyes have cone cells that perceive RGB three different colors, color in optics can be divided into “three primary colors” : Red, Green, and Blue. The wavelength range of each primary color is also different (approximate range) :
● R: 620-780 nm
● G: 500-560 nm
● B: 430-470 nm
02 Restore the color of the image
On the research and application of light and image, the camera industry emerges at the moment. However, camera Sensor (CCD and CMOS) can only accept the intensity of light, but cannot perceive the color of light, that is, cannot identify the frequency or band of light. Therefore, if CCD and CMOS are directly used, the output can only be gray image, as shown in the figure below:
In order to get the color image that our eyes are used to as much as possible, and output color to restore more real image as much as possible, people put forward two solutions:
Scheme 1: Bayer array scheme
This relatively low cost is proposed by foreign expert bayer:
By adding an RGB filter array like the one below in front of the sensor, each filter point can only transmit one color, and make each color filter point correspond to the lower pixel point one by one.
Through the above method, the above colorless gray image can be processed into the red, green and blue intensity map as shown in the following figure. It can be seen that the initial image is slightly green, which is not what people want.
Next is to guess color, that is, according to the gray values of a pixel and its surrounding red, green and blue, calculate the RGB of the pixel by interpolation. There are many interpolation algorithms. The simplest is to assign the color value of the adjacent pixel to the pixel, or assign the gray value of the color in the neighborhood to the pixel after averaging. There are many algorithms (refer to Bayer filter interpolation for specific algorithms).
According to the ideal situation, each pixel has 3 values, but in fact, R and B only have 1/4 respectively, and G has 1/2. Therefore, in the images obtained by bayer pattern, only 1/3 of the contents are actually true, and the rest are obtained according to the prior knowledge interpolation. This also shows that there is a lot of redundant information in natural images.
Scheme 2: Three filter schemes
This cost is relatively high, can ensure the degree of color restoration to a greater extent.
The principle is to use a splitter or prism to divide the light passing through the lens into three beams and then reach three sensors. Each sensor has a different filter corresponding to the RGB three colors, so as to ensure the accuracy of the colors collected by the camera.
01 Application camera
SONY FCB-EW9500H uses a 1/1.8 Starvis™ CMOS 4.17 megapixel image sensor with an effective pixel of about 4.17 million. It uses 7 petal aperture to reduce glare/halo and has excellent optical axis consistency to effectively restore the color of the image.