A CMOS Image Sensor Based On Intra-frame Redundancy Compression

Owing to the great amount data of output, CMOS image sensor can’t meet themulti-sensor system’s requirements of real-time transmission and storage, whichlimits the development of itself. Therefore, improving the CMOS image sensor outputefficiency and completing image compression steps in sensor process become one ofthe key problems in the world today. This paper conducts a thorough research mainlyin a CMOS Image Sensor based on intra-frame redundancy compression.This paper determines to make the discrete cosine transform of transform codingcompression method into the CMOS image sensor after researching the imageintra-frame redundant compression process and comparing with the variouscompression method. This paper designed a CMOS Image Sensor achievingTwo-Dimensional Discrete Cosine Transform with mixed signal processing. Utilizingthe row-column decomposition principle and the symmetrical transform kernel of2D-DCT, pixel array is divided into8×8block. Adding two groups of samplecapacitor arrays, register control the switch control module to choose the capacitorratio of corresponding transform kernel coefficient with analog accumulatorcompleted the column transform. Using MUX with pixel array completed columntransformation sign conversion. Switch capacitance amplifier and digital domainaccumulation complete the row transform. Then, the pixel array, multiplexer, analogaccumulator and switch capacitance amplifier are deeply analyzed and designed.Finally, the system architecture is modeling tested and verified and the circuit moduleis simulated in CHRT0.35μm standard CMOS process.Modeling results show that this CMOS Image Sensor architecture achieves2D-DCT accurately and the PSNR is73.54dB when retain15%DCT coefficientcompression. The number of computation periods is25%of the traditionalmixed-signal processing method. Circuit simulation results show that the circuitmodule meet the function and performance requirements of the system. Theinnovation point is that this paper presented a CMOS image sensor which cancomplete the2D-DCT in the light signal perception process, and achieve the purposeof parallel execution operation by changing the variable ergodic sequence, whichimprove the speed of traditional mixed signal processing method and satisfy any quantization coding needs. This CMOS Image Sensor improve the traditional CMOSImage Sensor output efficiency, suitable for high efficiency imaging system such asbiological and medical treatment, wireless sensor network and so on.