An Efficient FPGA Implementation of Optimized Anisotropic Diffusion Filtering of Images
An Efficient FPGA Implementation of Optimized Anisotropic Diffusion Filtering of Images
复制标题
DOI：
10.1155/2016/3020473
复制
发表时间：
2016-01-01
影响因子：
4.3
通讯作者：
Ghosh, Ranjan
中科院分区：
文献类型：
--
作者：
Pal, Chandrajit;Kotal, Avik;Ghosh, Ranjan
数字图像处理是一个令人兴奋的研究领域，具有各种应用，包括医疗，监控安全系统，国防和空间应用。作为预处理步骤的噪声去除有助于提高信号处理算法的性能，从而提高图像质量。Perona和Malik提出的各向异性扩散滤波可以用作边缘保持平滑器，去除图像的高频分量而不模糊其边缘。在本文中，我们提出了一个边缘保持各向异性扩散滤波器的数字图像的FPGA实现。所设计的架构完全取代了卷积运算，并使用内核内相邻强度的简单算术减法来实现卷积运算，在内核内并行执行多个运算。为了提高图像重建质量，对控制滤波过程的扩散系数参数进行了适当的分析。它的信号行为已经研究了随后的缩放和微分信号。所提出的设计的硬件实现表现出更好的性能，重建质量和加速性能方面的软件实现。相对于其他相关工作，它还减少了计算、功耗和资源利用。
Digital image processing is an exciting area of research with a variety of applications including medical, surveillance security systems, defence, and space applications. Noise removal as a preprocessing step helps to improve the performance of the signal processing algorithms, thereby enhancing image quality. Anisotropic diffusion filtering proposed by Perona and Malik can be used as an edge-preserving smoother, removing high-frequency components of images without blurring their edges. In this paper, we present the FPGA implementation of an edge-preserving anisotropic diffusion filter for digital images. The designed architecture completely replaced the convolution operation and implemented the same using simple arithmetic subtraction of the neighboring intensities within a kernel, preceded by multiple operations in parallel within the kernel. To improve the image reconstruction quality, the diffusion coefficient parameter, responsible for controlling the filtering process, has been properly analyzed. Its signal behavior has been studied by subsequently scaling and differentiating the signal. The hardware implementation of the proposed design shows better performance in terms of reconstruction quality and accelerated performance with respect to its software implementation. It also reduces computation, power consumption, and resource utilization with respect to other related works.