许博闻, 王敏, 李琛, 王鹏飞, 王国兴. 基于图像去噪的导向滤波算法的硬件实现[J]. 微电子学与计算机, 2019, 36(7): 22-26.
引用本文: 许博闻, 王敏, 李琛, 王鹏飞, 王国兴. 基于图像去噪的导向滤波算法的硬件实现[J]. 微电子学与计算机, 2019, 36(7): 22-26.
XU Bo-wen, WANG Min, LI Chen, WANG Peng-fei, WANG Guo-xing. Hardware design for image denoising using guided filter[J]. Microelectronics & Computer, 2019, 36(7): 22-26.
Citation: XU Bo-wen, WANG Min, LI Chen, WANG Peng-fei, WANG Guo-xing. Hardware design for image denoising using guided filter[J]. Microelectronics & Computer, 2019, 36(7): 22-26.

基于图像去噪的导向滤波算法的硬件实现

Hardware design for image denoising using guided filter

  • 摘要: 在图像去噪算法领域, 导向滤波算法在去噪的同时能够保持更多的图像细节, 起到很好的边缘保持效应.相比于其他边缘保持算法, 导向滤波算法运算速度更快, 在CMOS图像传感器领域, 能更快更有效的对图像进行去噪处理.本文在导向滤波原有的算法公式上, 提出一种改进的硬件实现方法, 并应用于图像传感器芯片中, 像素分8通道进入模块进行处理, 采用RGB色彩模式, 算法模块最高时钟频率为100 MHz, 每行最大处理4 800个像素, 一共占用46.875 kB片上存储, 以及XCRU040的13%的LUTS资源.

     

    Abstract: In the field of image denoising, the guided filter can reserve more image details and has a good edge-preserving effect. Comparing with other edge-preserving algorithm, the guided filter presents a higher efficiency. As a result, in the field of CMOS Image Process(CIS), the hardware design of the guided filter is able to process image denoising quickly and effectively. In this paper, based on the original algorithm of guided filtering, a hardware implementation method is proposed and used in the CIS. The pixels are divided into 8 channels to enter module for processing. Color mode is RGB. The highest clock frequency is 120 MHz and the pixels number processed per line by the module can reach 4 800. The total amount of the on-chip memory occupied by this module is 46.875 kB.Total LUTS occupy 13% resource of XCKU040.

     

/

返回文章
返回