Proc. SPIE. 8908, International Symposium on Photoelectronic Detection and Imaging 2013: Imaging Sensors and Applications
KEYWORDS: Polarization, Image processing, Digital signal processing, Parallel processing, Image registration, Logic, Field programmable gate arrays, Control systems, Telecommunications, Signal processing
Compared with traditional optical intensity image processing, polarization images processing has two main problems.
One is that the amount of data is larger. The other is that processing tasks is more complex. To resolve these problems,
the parallel processing system of multi-channel polarization images is designed by the multi-DSP technique. It contains a
communication control unit (CCU) and a data processing array (DPA). CCU controls communications inside and outside
the system. Its logics are designed by a FPGA chip. DPA is made up of four Digital Signal Processor (DSP) chips, which
are interlinked by the loose coupling method. DPA implements processing tasks including images registration and
images synthesis by parallel processing methods. The polarization images parallel processing model is designed on multi
levels including the system task, the algorithm and the operation. Its program is designed by the assemble language.
While the polarization image resolution is 782x582 pixels, the pixel data length is 12 bits in the experiment. After it
received 3 channels of polarization image simultaneously, this system implements parallel task to acquire the target
polarization characteristics. Experimental results show that this system has good real-time and reliability. The processing
time of images registration is 293.343ms while the registration accuracy achieves 0.5 pixel. The processing time of
images synthesis is 3.199ms.