Dynamic range of image sensors
What is dynamic range?
Dynamic range is one of the key parameters of an image sensor. It defines the range of light intensity values the sensor can detect, from shadow detail to highlight detail, and thus determines the captured image's detail, tonal range, and features.
Film photography
Based on photochemical principles, during camera exposure light passes through the lens and reaches silver halide crystals on the film, causing a change in the film's optical density. The relationship is nonlinear: as exposure increases, optical density changes accordingly. When light hits the silver halide, silver is converted to metallic silver. After development and fixing, this becomes the black-and-white image on the film base. Color film uses three layers of silver halide emulsions to represent the primary colors.
Digital imaging
Image sensors such as CCD and CMOS sample the incident light with photosensitive elements, producing thousands of pixels whose responses are proportional to exposure. These are converted to analog voltages, then digitized by an A/D converter into digital signals, and finally transformed by a microprocessor with nonlinear processing into standard image storage formats such as BMP, JPEG, and TIFF on physical media.
Definition of dynamic range
Dynamic range is defined as the ratio between the maximum measurable light intensity and the minimum measurable light intensity. It can also be defined as the ratio between the number of electrons at pixel full-well capacity and the number of noise electrons under dark conditions. It is typically expressed in decibels (dB):
In natural scenes, the dynamic range from a moonless night to direct sunlight is roughly 180 dB, while conventional image sensors are usually limited to around 70 dB due to well capacity constraints.
When the same scene exhibits a wide variation in luminance, it is referred to as high dynamic range (HDR). Conversely, images with a limited luminance range are considered low dynamic range (LDR).
The imaging process of a camera essentially maps a high dynamic range real-world scene to a lower dynamic range image using a nonlinear mapping, for example:
Factors affecting dynamic range
Well capacity
Conventional image sensors have a limited photocharge well capacity in each photosite, so the amount of stored charge is limited. When illumination exceeds a certain level, the stored charge reaches saturation and the sensor can no longer record additional photons. Once these pixels are saturated, charge may overflow and cause information loss.
For example, highlight overflow may cause pixels near a saturated red region to read 255 even though their true values are below 255, resulting in lost detail in the highlights. If exposure time is reduced to prevent highlight clipping, many pixels in dark regions do not have enough time to collect photons and may read 0, causing loss of shadow detail.
Bit depth
Most current output and display systems are constrained by software and hardware to 8-bit processing, representing only 256 gray levels. This limits the number of distinguishable luminance levels compared with real scenes.
Therefore, regardless of how large the input luminance range is, if the image is digitized to 8-bit, it can only represent 256 gray levels. Sensors with different dynamic ranges map those 256 levels to different actual luminance spans: the larger the dynamic range, the better the sensor can represent natural luminance variations.
Methods to increase dynamic range
Adjusting well capacity
Increasing pixel full-well capacity or reducing pixel noise are direct ways to increase sensor dynamic range. However, modern pixels are shrinking in size and have fixed fill factors for a given pixel architecture, making it difficult to extend dynamic range significantly by increasing full-well capacity alone.
Increasing bit depth
Raising image bit depth is a common approach. Increasing the number of gray levels allows more tonal steps and a wider apparent dynamic range in the digitized image.
Logarithmic response techniques
Modifying the pixel output response curve can delay the onset of pixel saturation, effectively increasing the pixel charge-handling range and extending dynamic range.
Multiple exposure techniques
Without changing the photodiode or pixel architecture, traditional low dynamic range sensors can produce high dynamic range images by combining multiple exposures taken at different exposure values.
