CHEN RUI KE Electronic Company Limited

The minimum value of S for an image frame is 1, and the maximum value is limited by the data bandwidth, that is, the maximum value that can be stored in the data frame. An 8-bit grayscale image, the maximum value of 255, then the theoretical dynamic range is 48dB, 10-bit image dynamic range of 60dB, 20-bit image dynamic range of 120dB. In fact, dynamic range is a general concept, different signals or variables can define their own dynamic range, image sensors have dynamic range, monitors, projectors, printers and so on have their own dynamic range. We can even define a person's dynamic range, if this person can endure hardship when conditions are hard, enjoy when conditions are good, and can both make do with and pay attention to, this is a high dynamic range ruthless person. The dynamic range of the image and the dynamic range of the scene are, for the most part, at odds. The signal S of the scene is not the grayscale value of the image, it is the brightness of the emitted light of the scene. It can be understood that S Max is the brightness of the brightest part of the scene, and S min is the brightness of the blackest part of the scene. It is related to but not equal to the image dynamic range. At the same time, image post-processing often compresses linear data into nonlinear output, which also magnifies differences in image numerical values and scene dynamic range. If the brightness of the scene is taken as the abscissa, and the output data of the image sensor is taken as the ordinate, the image sensor collects and maps the scene with a certain brightness range to its output. The actual dynamic range of the image sensor is usually lower than the dynamic range of the scene, and the ability of the sensor can only collect the brightness range of the scene in the horizontal coordinate corresponding to the red box. The position of the red box needs to be dynamically adjusted and moved to adapt to the change of the scene brightness, which is the task of the Auto Exposure module in the imaging algorithm. The first one is to dynamically change the sensitivity of the pixel to extend the dynamic range. The mapping of the image sensor to the scene brightness becomes nonlinear. With the increase of the environment brightness, the pixel sensitivity gradually decreases, and the sensitivity changes from a linear function of the brightness to a piecewise function. Charge accumulation is divided into three segments, high sensitivity at low brightness, corresponding to black charge, then also medium sensitivity at medium brightness, corresponding to blue charge, and finally lowest sensitivity at highest brightness. It can be seen from the coordinate figure that at this time, the potential well capacity of the pixel, that is, the ordinate, does not increase, but the mapped scene brightness range, that is, the horizontal coordinate, can be significantly increased, achieving the goal of increasing the dynamic range. Onsemi's automotive image Sensor product line launched a 300k pixel variable sensitivity sensor in the early stage, which is based on this technology. The biggest challenge of this technology is that it changes the sensitivity characteristics of the pixel, so that the sensitivity of the linear characteristics becomes nonlinear. The limited dynamic range extension capability can only be barely used for sensors with large pixel black and white images. At present, this kind of technology has been gradually eliminated from the market. The second high dynamic range technique is time division multiple exposure. This is the technology used by mainstream image sensors in cars. The way to do this is that the image sensor changes the exposure time and continuously exposes multiple frames of images, and then selects the appropriate pixels from them to merge into a frame image. The sensor changes the exposure time, which is equivalent to its own automatic exposure function, and samples the scene with different brightness to obtain multiple red boxes, and then concatenates the dynamic range. The advantages of this technique are: the capacity of the pixel potential well need not be increased, only the data bandwidth needs to be increased; The duration control of each exposure can be very accurate, and the final fitting image brightness has good linearity. Dynamic range extension is easy and 140dB dynamic range can be achieved only with time division technology. Time and multiple exposure technology has an intractable problem. Because the continuous exposure time of the Sensor is successively lagged, when there are fast moving objects in the scene or severe illumination changes such as LED strobe, moving object artifacts and color noise will appear after multi-frame image fitting. ADAS algorithms need to be trained specifically for this kind of noise.