CCD vs. CMOS: Image Artifacts to Consider with CMOS Image Sensors

The high-resolution and frame speed combination with CMOS image sensors is very attractive and the major driving factor towards the move to CMOS with industrial cameras.  A consideration with CMOS image sensors are the different types of defects present and whether the camera manufacturer corrects for them. 

We investigated this technology thoroughly in order to approach CCD image quality with CMOS image sensors as much as possible. Some well know defects are relatively easy to correct but some are more challenging defects, such as those dependent on temperature.

The better known defects are:

• Pixel defects:  single, clusters, rows, columns
• DSNU (Dark Signal Non Uniformity)
• PRNU (Pixel Response Non Uniformity)

Some lesser understood defects include column noise, line noise, and black level shift (see images below).  These are more difficult to correct because they are image content dependent.  

            Column Noise                              Line Noise                          Black level shift

Defect pixel correction and flat field correction are examples of corrections that can best be implemented right in the camera.

Just like in almost every situation it is best to eliminate problems at the source. It reduces noise and other problems further in the process. In this case correcting artifacts in the camera has the following advantages:

– the camera manufacturer has intimate knowledge of the sensor

– in the camera you have more bits available

– in the camera you have more information available (like temperature)

– it saves processing power – system costs (development + BOM)

For instance, new pixel defects will become visible with time – so it is advantageous if new defect maps can be uploaded to camera.  Or, DSNU is temperature (and gain) dependent, so corrections must be refreshed regularly.

High end (metrology) applications need these kind of corrections to achieve the precision and accuracy required.