Graphics Reference
In-Depth Information
3.4.2 Imaging Sensors
Driving force behind the perfection of CMOS-based imagers were high volume and
relatively short product life cycle applications such as mobile telephones, tablets,
and PCs. The technical specifications required by automotive suppliers for imaging
sensors targeted toward ADAS market are different and very application specific.
High dynamic range and low light performance, wide operating temperature range,
fast motion, support for functional safety, and device cost are five most important
parameters for imaging sensors in ADAS.
The requirements for an automotive imaging sensor are challenging as device
needs to produce reliable image output under rapidly changing illumination level
conditions at extreme temperatures.
The human eye has a wide dynamic range of about 200dB. The eye has three
mechanisms to achieve such a wide dynamic range: two types of photoreceptor
cells (rods and cones) with different photosensitivities, logarithmic photoreceptor
response, and shift of the response curve according to the ambient light level. The
front camera system must exhibit extremely high optical dynamic range to capture
scene details and detect objects in very bright and very dark parts of frame (
115dB
for front view camera applications). To achieve Wide Dynamic Range (WDR), auto-
motive imaging sensor suppliers typically use different techniques [
12
,
13
] such as:
skimming WDR, staggered multicapture WDR, down-sampling WDR, split pixel
WDR, and log sensors.
Although imaging sensors with global shutter offer elimination of motion artifacts
through simultaneous capture of the entire frame, this advantage comes at the expense
of increased pixel noise and reduced light sensitivity. Today majority of automotive
imaging sensors use a rolling shutter and algorithms are designed to deal with motion
artifacts.
Unlike the human eye, all imaging sensors are monochromatic. To obtain the color
information, a color mosaic filter is placed over the sensor pixels, which typically
cuts spatial resolution of a color sensor in half compared to grayscale. As they offer
nearly double the spatial resolution of a color sensor, grayscale sensors are much
more sensitive to variations in brightness.
Development life cycle timeline of ADAS sets pace for automotive imaging sen-
sor road maps. Typically automotive imager resolution increases every three to four
years. Downside of frame resolution increase is lower sensitivity to light due to
reduction of pixel size. To combat reduced pixel light sensitivity imaging sensor sup-
pliers are introducing 'clear pixels' without color filter. In addition to monochrome
and Bayer color pattern, the imaging sensors in ADAS support different arrange-
ments of color filters, such as RCBC (Red-Clear-Blue-Clear), RGBC (Red-Green-
Blue-Clear), RGBIr (Red-Green-Blue-InfraRed), and RCCC (Red-Clear). Typically
a specialized ISP is required to convert this raw output from an imaging sensor to
a format that can be used by embedded vision algorithms. Low-resolution imaging
sensors typically integrate the ISP functionality on the die.
>
