A novel hardware architecture of Decision Based Adaptive Denoising Algorithm for removing salt & pepper noise

Abstract

Smart Visual Internet of Things (SVIoT) has been widely used in different industries due to the rapid growth in wireless sensor devices with faster computation capability and high-speed connectivity. Healthcare, smart cities, cloud computing, artificial intelligence, and the automotive industries are supported by the IoT platform [1]. The SVIoT based devices consist of various components such as sensors, cameras, GPS etc. In real time applications, Image is acquired by multiple sensor devices. Generally, captured images suffer from impulse noise during acquisition, processing, transmission and storing, which lead to loss of information and degraded performance of the devices [2].

Generally, noise suppression techniques are commonly used to remove unwanted noise while preserving the original data in digital images. Digital images and video signals are often corrupted by salt & pepper noise in the process of image and video signal transmission and acquisition. However, the high-speed and efficient noise removal techniques are required for various image & video applications. The conventional noise removal filters are done for the elimination of noise from the digital images. The standard digital image processing necessary to remove impulse noise without disturbing the information at edges. The conventional linear filtering methods fail when the noise is ineffective and non-additive in impulse denoising. This has led the new research to the use of non-linear filter techniques [3]. A small size window provides low efficient information for the denoising on the converse a large window size is better efficient for removing salt & pepper noise, but the image is blurred. However, the proper trade-off in choosing the size of window before implementing an efficient denoising algorithm. A widely used non-linear filter is median filters. Median filters were found to have good efficiency to remove impulse noise as well as preserve the edges. Mostly, median filters are preferred for removing salt & pepper noise from images. 

Median filter is a state-of-the-art approach for removing salt & pepper noise in image processing applications. The main drawback of Standard Median filter is that the pixels in the entire window on the digital image are replaced with median value irrespective of whether the neighborhood pixels are corrupted or not [4].

A new Decision Based Adaptive Denoising Algorithm (DBADA) and hardware architecture is proposed for restoring the digital image that is corrupted by impulse noise. The adaptive median filter is modified in the proposed DBADA. The proposed DBADA detects only the corrupted pixels, and that pixel is restored by the noise-free median value or previous value based upon the noise density in the image. The proposed DBADA uses a 3 × 3 window initially and adaptively goes up to a 7 × 7 window based on the noise corruption by impulse noise in the current processing window. The proposed architecture was found to exhibit better visual qualitative and quantitative evaluation based on PSNR, IEF, EKI, SSIM, FOM, and error rate. The DBADA architecture also preserves the original information of digital images with a high density of salt & pepper noise, when compared to many standard conventional algorithms. The proposed architecture has been simulated using the VIRTEX 7 FPGA device and the reported maximum post place and route frequency is 112 MHz.