【摘要】 为了实现CMOS图像传感器(CIS)片上系统(SoC)中伽玛(γ)校正的低功耗设计,同时又保证校正的精度,提出一种查找表和直线拟合相结合的γ校正技术。算法对灰度值较低的像素使用直接查找表方法校正,对于γ曲线上升缓慢部分的像素采用分段直线拟合的方法。在直线分段时,使用外层分段与内层分段相结合的方法,达到了分段优化的目的。算法保证了图像校正精度,与使用完全查找表法相比,误差在0.5 pixel之内。基于该方法设计了一个8 bit输入/8 bit输出的VLSI模块,通过FPGA对模块进行了验证,模块占用723个LE和195个LC寄存器,比完全查找表法减少了硬件资源耗费,实现了低功耗设计。系统最大工作频率可达148 MHz,完全满足实时处理的需求。更多还原

【Abstract】 In order to implement the low-power design of Gamma correction for image signal processing circuits of CMOS image sensor SoC and ensure the calibration accuracy at the same time,a new gamma correction method is presented.The proposed algorithm combines the look-up table and linear fitting to correct images.To ensure the accuracy,this method uses the look-up table to emend low-value pixels.For high-value pixels,a method of combining look-up table and linear fitting is put forward.When linear fitting,the outer and inner sub-section methods are used to achieve sub-optimal.Simulation results with real images proved that the propose method is effective.A design of 8 bit input/8 bit output is implemented with VLSI technique,and in order to achieve the design flexibility,the registers that save lines′ slopes can be accessed by external user,which can amend gamma values for different application requirements.By FPGA validation,the module takes up 723 LE and 195 LC registers,which reduces hardware resource consumption than full look-up table method.This module can work with maximum frequency of 148 MHz,wihch can meet the real-time processing requirements.更多还原