CFA图像插值算法在DSP实时系统上的应用  

Application of CFA images interpolation algorithm in DSP real-time system

作　　者：孙海江[1,2] 吴川[1] 王延杰[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院研究生院,北京100049

出　　处：《红外与激光工程》2011年第6期1187-1192,共6页Infrared and Laser Engineering

基　　金：中国科学院创新项目(O60Y32T060)

摘　　要：彩色滤波阵列(CFA)格式的图像需要进行图像插值,才能获得全彩色图像。但是,由于相机噪声的存在,不仅该点的彩色信息失真,同时还会影响邻近像素利用该点信息进行插值的图像颜色。为了得到高质量的全彩色图像,采用了一种基于梯度的滤噪插值算法,在保留了图像边缘和细节信息的同时,滤除了噪声,使图像的信噪比提高了近1倍;由于这种数字相机没有白平衡处理,使插值恢复的彩色图像颜色与实际景物相比存在较大差异,为了消除这种差异,实现了自适应白平衡校正方法,使图像方差提高了近20%,与人眼观察到实际景物的颜色更接近;整个彩色编码算法应用了DSP的并行流水线技术,保证了其在DSP上运行的实时性,处理频率达到25 Hz。The images of color filter array（CFA） form need an interpolation process commonly for producing the full-color images.The noise of imaging sensors not only corrupts the color filter array,but also influences the quality of images at the same time.In order to acquire high quality full-color images,a sort of viable and effective interpolation algorithm was adopted based on gradient,at the time of reserving image border and detail information clearly,the noise was removed and video SNR was nearly two times than before.Because of not having white balance in this digital camera,it has many difference from the color of images based on interpolation algorithm to real scene.In order to eliminate the difference,adaptive adjust for color balance was realized,image mean square was nearly increased by 20%,and it was more nearness for human eyes to look scenery.The color filter method for CFA images by parallel pipeline can solve the real-time problem on DSP and disposing frequency can reach 25 Hz.

关 键 词：彩色滤波矩阵 插值算法 自适应白平衡 DSP 并行流水 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]