# Does random access to image sensors require extra hardware?

Jun Ohta's "Smart CMOS image sensor and applications" says that random access to pixel arrays is possible. Using Row parallel readouts and column decoders, one can select which pixels to readout and which not. Since the sensor outputs only one pixel at a time, it seems to me that the sensor has all the necessary hardware to output a rectangular window, which is a sub-image of the total sensor.

Am I right or does the sensor need extra hardware to achieve a "random capture"?

• why would you select pixels that you do not want to read out? – jsotola Mar 13 '18 at 8:22
• your reasoning is flawed Since the sensor outputs only one pixel at a time, it seems to me that the sensor has all the necessary hardware to output a rectangular window, which is a sub-image of the total sensor. ... it has the hardware to output the full frame ... you cannot assume that you can choose a sub-image – jsotola Mar 13 '18 at 8:28
• You can read out small regions of interest (ROI) in CMOS cameras and this usually done to increase the framing rate. The bandwidth to the outside world is limited (i.e. USB connection) and to get a faster rate a smaller image is required. Obviously, this isn't done for low quality webcams. Google: CMOS camera region of interest – D Duck Mar 13 '18 at 9:05
• Tracking incoming air-air missiles, ehh? You'll need control over at least the row-select hardware, and preferably the ability to specify the starting column to read out, and how many columns to read out. Also you'll need to erase and restart the exposure time. – analogsystemsrf Mar 13 '18 at 17:59
• @analogsystemsrf I agree with the former part, i.e., I'll need control over the row-select and column decoder hardware; the latter, i.e., the erase and restart, does not sound like a necessity, though. Can you explain a bit more? – Ekdeep Singh Lubana Mar 14 '18 at 7:50