FPGA Implementation of Modified Recursive Box Filter-Based Fast Bilateral Filter for Image Denoising
- PDF / 1,496,402 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 47 Downloads / 214 Views
FPGA Implementation of Modified Recursive Box Filter-Based Fast Bilateral Filter for Image Denoising Gollamandala Udaykiran Bhargava1 · Sivakumar Vaazi Gangadharan2 Received: 1 July 2019 / Revised: 28 August 2020 / Accepted: 4 September 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Bilateral filter (BF) is a type of edge-preserving smoother that is widely utilized in most image processing, computational photography, and computer vision applications. This article presents FPGA implementation of a modified recursive box filter (MRBF)based fast BF (FBF) architecture for reducing computational complexity, area, and power. In addition, implementation of MRBF uses a modified carry select adder using quantum-dot cellular automata technology that offers lower area, less power consumption, and less delay. Furthermore, as an application, an investigation of image denoising with the proposed MRBF-FBF is also performed. Extensive simulation and hardware results disclose that the proposed FPGA implementation of MRBF-FBF outperforms the conventional filtering techniques in terms of quality assessment of the denoised image, such as peak signal-to-noise ratio and structural similarity index. Keywords FPGA · Quantum-dot cellular automata · Image denoising · Box filter · Recursive box filter · Modified recursive box filter · Bilateral filter · Carry select adder · Modified carry select adder
1 Introduction Original image reconstruction from its noisy visual aspect is a primary stage of most of the applications in image processing and computer vision. Usually, noise is introduced during the steps of image acquisition, image transmission or image reception.
B
Gollamandala Udaykiran Bhargava [email protected] Sivakumar Vaazi Gangadharan [email protected]
1
Electronics and Communication Engineering, Sathyabama Institute of Science and Technology, Jeppiaar Nagar, Rajiv Gandhi Salai, Chennai, Tamilnadu, India
2
Electronics and Communication Engineering, Vidya Jyothi Institute of Technology, Hyderabad, Telangana, India
Circuits, Systems, and Signal Processing
Frequently, these noise attributes reckon on the kind of sensors, exposure duration, dimension of pixel, speed of ISO, ambient illumination level and temperature. Practically, additive white Gaussian noise (AWGN) is a kind of noise that affects most of the applications in image processing. Let y be the noisy or observed image, z is the latent image and the zero mean Gaussian noise is denoted as η ∈ ℵ 0, σ 2 with σ 2 variance. Then, y can be derived from y z+η
(1)
There are three different classifications in available denoising approaches, namely spatial, spectral, and dual domain. The fundamental thought behind the approaches of spatial filtering dissent only to the weights or kernels magnitude those are computed either locally or no-locally to approximate unlike data elements in an image [22]. BF [31], non-local means (NLM) filter [6] and the guided image filter [17] are some of advanced edge-preserving filter
Data Loading...
