Study On CMOS Image Sensor Based On Compressed Sensing

ABSTRACT:As a new signal sampling theory, CS (Compressed Sensing) makes the technology problem of the information acquisition be transferred from the sampling end to the receiving end, and signals with low sampling rate can be accurately reconstructed. So it has been widely concerned by experts and scholars from all over the world. Compared to the theory of CS which is constantly improved, the implementation of the hardware structure is a new direction. In the paper, the CS-based CMOS image sensor (CIS) is treated as a study object, and advantages and disadvantages of existing implementation methods are summarized. The sensor is studied for the purpose of simultaneously sensing and compressing the data before analog-to-digital conversion, and a specific hardware implementation is proposed.By analyzing structures and optical signal acquisition principles of traditional pixel circuits, a CS-based current-mode CMOS active pixel sensor with storage and multiple selective outputs is proposed. The optical signals can be output only once in traditional pixel structures, so it can’t be implemented multiple computations for a signal and a measurement matrix based on CS. For the pixel with storage function, the global shutter of CIS can be implemented. Further, an optical signal can be stored in the active pixel sensor, so the signal in the pixel can be selectively output many times. That is to say, the same signal can be computed with a measurement matrix many times.By summarizing the existing implementation methods of CS-based CIS, a hardware implementation of CS-based CIS is proposed, which senses and compresses signals at the same time using the global shutter approach. The sensor includes a pixel array with storage function, a pseudo-random sequence generator, a random sequence store and control unit, a timing control unit and per-16-columns trans-impedance amplifiers.The CS-based global shutter CIS reduces data acquisition amount, lowers requests on the sampling speed of an analogue-digital converter and saves data storage space on the premise of guaranteeing that the resolution is as same as that of the traditional method as image sensing and compression is finished before analogue-digital conversion based on a compressive sensing theory, improves the processing speed as a hardware way rather than a software way after the analogue-digital conversion is adopted for data compression, and can acquire transient images of an object moving at a high speed as a global shutter way is adopted.