Blog_banner_beyond_the_datasheet

Camera Dynamic Range Requirements for Machine Vision versus Visual Applications

Posted by Gretchen Alper on Tue, Oct 29, 2013

In photographic terms, dynamic range indicates the ability of a camera to reproduce the brightest and darkest portions of an image.   It is one of the key parameters for determining the right machine vision or industrial camera for your system [for more background information on using dynamic range to compare cameras, click here].  But there is more to dynamic range beyond the value provided on the specification sheet - there are different situations with varying requirements such as machine vision or viewing applications. Even in machine vision (controlled lighting situations), there can be different needs depending on the system environment.  There is functionality that can be added to the camera for better performance in specific conditions.  Here are more details…

Dynamic range can be linear or non-linear or even multi-slope linear.  Typically for machine vision applications you want a linear dynamic range, as you need constant and predictable behavior over the entire image for accurate and repeatable measurements. This is even more so for metrology methods that use color in the images. Since CMOS and CCD image sensors are inherently non-linear in response (here non-linear means the deviation from the linear response curve), it is important that the camera controls the image sensor and outputs a linear response. The linearity of the camera response results in a higher accuracy of the algorithms used in the metrology method. The reason for this is that most of the mathematics used in the algorithms are based on linear calculations and transformations.

For viewing applications (humans rather than computers using the images in often uncontrolled lighting environments), linearity is much less important.  What is more important is that the camera responds to light in a similar way to the human eye.  Human eyes have a non-linear, more logarithmic response to light intensity.  For example, if there were twice as much light in a given scene then the resulting image from a camera with a linear response would be two times as bright.  A human, however would see the scene as incrementally brighter but not twice as bright. 

Some cameras offer the option to have both (linear and non-linear dynamic range) for optimal use in different situations. This non-linear range is often a multi-slope linear response. There can be an extended dynamic range function embedded in the camera by introducing a non-linear, multi-slope response.  This can result in a dramatic dynamic range increase such as 60 dB to 90 dB. 

Viewing applications are often in outdoor environments with high dynamic range scenes.  This function (HDR) can allow for details in both the high lights and in the shadows of an image.  This is different from the HDR feature available in many smart phone cameras as those use essentially linear sensors but combine pieces from multiple images to produce a higher dynamic range image. This smart phone implementation is therefore less useful when you have a moving subject in the scene.

Also, with machine vision, there are times when the extended dynamic range function can be beneficial.  For instance, when you are concerned with one end of the light intensity curve or the other.  In light-starved systems (read noise limited), you need a linear response in the dark portion while maximum Full Well is of less importance.  In systems with plenty of light (shot noise limited), you want a linear response in a brighter portion of the scale but without saturation of the image.  This allows you to still have linear dynamic range but just for the conditions your system operates in.  The camera/image sensor can be optimized automatically with the HDR function in this way to have greater details without affecting repeatability.  Figure 1 shows an example of linear fit to the response curve for very low light intensity using the HDR function.  This would provide higher linear dynamic range for systems with low light conditions.

 

HDR LR graph 

Figure 1.  Linear Response (LR) Versus HDR

 

As an example of the improvements possible, here is a scene that requires a lot of dynamic range:

Adimec dynamic range off

 

And here is the same image taken with the HDR extended dynamic range mode enabled which is a feature available in the Adimec Quartz Q-4A180.  

Adimec dynamic range HDR on

 

describe the image

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Topics: Image Quality Improvements

Previous blog:

Next blog: