A GPU-Accelerated Real-Time NLMeans Algorithm for Denoising Color Video Sequences

Abstract

Abstract. The NLMeans filter, originally proposed by Buades et al., is a very popular filter for the removal of white Gaussian noise, due to its simplicity and excellent performance. The strength of this filter lies in exploiting the repetitive character of structures in images. However, to fully take advantage of the repetitivity a computationally extensive search for similar candidate blocks is indispensable. In previous work, we presented a number of algorithmic acceleration techniques for the NLMeans filter for still grayscale images. In this paper, we go one step further and incorporate both temporal information and color information into the NLMeans algorithm, in order to restore video sequences. Starting from our algorithmic acceleration techniques, we investigate how the NLMeans algorithm can be easily mapped onto recent parallel computing architectures. In particular, we consider the graphical processing unit (GPU), which is available on most recent computers. Our developments lead to a high-quality denoising filter that can process DVD-resolution video sequences in real-time on a mid-range GPU.

Thoracic low-dose CT image processing using an artifact suppressed large-scale nonlocal means

“…In the processing of LNLM, all threads in the block-grid structure execute simultaneously to perform all the pixel-wise operations, which include the pixel-wise 1D nonlinear diffusion in (4)- (8), the pixel-wise weight calculations for each patch pairs in (13), and the pixel-wise ˆi f calculations in (10). We also further reduced the computation reduction by applying the parallelization optimization in [26][27]. Runtime comparison in the practical experiments indicates that the parallelized operation is more than 100 times faster than the previous serial version.…”

BM3D image denoising using heterogeneous computing platforms

“…The NL-means algorithm has been proven to be well suited to GPGPU computation. Several implementations of NL-means using GPGPU have emerged in the field of biomedical image processing, where the algorithm has been extended to support also multidimensional denoising of magnetic resonance images or video sequences [30,31]. With a small search window the computational complexity is low and it is possible to implement the NL-means filter by using solely deterministic calculation.…”

Video Processing on GPU: Analysis of Data Transfer Overhead

“…Early research on GPU implementations of video and image processing, however, have mainly focused on accelerating their computational components, such as filtering [1], [2], encoding/decoding [3], [4], or analysis [5]. When the whole application is taken into account, the parallelization scheme has to consider that GPUs belong to a heterogeneous system.…”

“…Recently, the high computational power and low cost of GPUs make them attractive platforms to speed up video and image processing applications [1], [2], [3]. These applications are highly suitable to the massively parallelism offered by GPUs, where each pixel can be processed by a separate thread.…”