FPGA-based Research On Color Restoration Technology Of Bayer Image

Due to constraints of size and cost,camera manufacturers mostly choose onesensor structure.So that the final image of such camera is a Bayer image which composes three color blocks(red,green and blue).Compared with commercial cameras,the output of industrial cameras is RAW Bayer images without any processing,which is suitable for high-quality image processing.In order to ensure the output of a full-color image,the corresponding interpolation algorithm must be used,which can add the other two color components to each Bayer image pixel.This process is called demosaic processing.Because of the special scenes(such as military,aerospace,etc.),the output of industrial cameras need to have high frame rate.And then,real-time requirements are put forward for Bayer image color restoration.The existing algorithms can be mainly divided into two categories,namely classical interpolation algorithm and high-order interpolation algorithm.The classic interpolation algorithm is simple and convenient,and it is also easy to meet real-time requirements of the imaging system.Although the higher-order algorithm has higher color reproduction degree and more accurate color information,its calculation process often adds a large number of multiplication/convolutional calculations,which results in a big computational complexity and also hard to implement in hardware.In view of the above situation,this paper proposes a demosaicing algorithm which can accurately restore image color and minimize the complexity.Firstly,the algorithm preprocesses the target image,that is,Expands the image and use a unified algorithm to restore the entire image to reduce the amount of computation of channel of pixel at the image edge.Then the improved gradient interpolation is used to calculate the horizontal edge gradient value and vertical edge gradient value in a 5*5 neighborhood.Then,the direction of border is arbitrated by counting the sum of the two direction gradient values.Finally,the interpolation of G component is used to restore the other two components,which can greatly reduce the distortion of color introduced by the conventional algorithm.After system simulation,it is found that the proposed algorithm has good performance.While decreasing the amount of computation,it can greatly reduce the zipper effect and purple fringing phenomenon caused by classical algorithms.In order to realize the hardware implementation of the improved algorithm system,a set of image processing system based on FPGA is built in this paper.It involves a multi-device simultaneous access for DDR3 SDRAM,which needs to use FPGA to arbitrate DDR3 SDRAM.The use of this arbitration mode can ensure the order and efficiency of multi-device access,and greatly improve the efficiency of the system.Finally,the design is a hardware system,which based on a Virtex-6 FPGA,consisting of image acquisition,process and output functions,and realizes the process of storing the camera's raw data in SDRAM.And then,the color reduction of Bayer image is accomplished in FPGA.Furthermore,depending on the requirements of subsequent transmission,the output can be performed via Camera Link interface or DVI interface.A hardware system is realized which integrating CMOS camera image acquisition,Bayer image color restoration and color image output.The system uses Virtex-6 FPGA as the hardware core,and a rich external expansion interface is designed,which has the advantages of high reliability,real-time performance and easy expansion.