Taxonomy of Vectorization Patterns of Programming for FIR Image Filters Using Kernel Subsampling and New One

Abstract

This study examines vectorized programming for finite impulse response image filtering. Finite impulse response image filtering occupies a fundamental place in image processing, and has several approximated acceleration algorithms. However, no sophisticated method of acceleration exists for parameter adaptive filters or any other complex filter. For this case, simple subsampling with code optimization is a unique solution. Under the current Moore’s law, increases in central processing unit frequency have stopped. Moreover, the usage of more and more transistors is becoming insuperably complex due to power and thermal constraints. Most central processing units have multi-core architectures, complicated cache memories, and short vector processing units. This change has complicated vectorized programming. Therefore, we first organize vectorization patterns of vectorized programming to highlight the computing performance of central processing units by revisiting the general finite impulse response filtering. Furthermore, we propose a new vectorization pattern of vectorized programming and term it as loop vectorization. Moreover, these vectorization patterns mesh well with the acceleration method of subsampling of kernels for general finite impulse response filters. Experimental results reveal that the vectorization patterns are appropriate for general finite impulse response filtering. A new vectorization pattern with kernel subsampling is found to be effective for various filters. These include Gaussian range filtering, bilateral filtering, adaptive Gaussian filtering, randomly-kernel-subsampled Gaussian range filtering, randomly-kernel-subsampled bilateral filtering, and randomly-kernel-subsampled adaptive Gaussian filtering.

Effective Implementation of Edge-Preserving Filtering on CPU Microarchitectures

“…For referring LUTs, the set or gather SIMD instructions were employed. The outermost loop was parallelized by multi-core threading, and we had pixel-loop vectorization [50]. This implementation was found to be the most effective [50].…”

“…The outermost loop was parallelized by multi-core threading, and we had pixel-loop vectorization [50]. This implementation was found to be the most effective [50]. Notably, a vectorized exponential operation is not implemented in these CPUs.…”

Application of Fast Non-Local Means Algorithm for Noise Reduction Using Separable Color Channels in Light Microscopy Images

“…In addition, the search window and kernel sizes are closely related not only to image characteristics, but also to the calculation time (i.e., time resolution) [ 46 , 47 ]. Figure 6 shows the time resolution results of various search window and kernel sizes.…”

Design and Implementation of 2D Convolution on x86/x64 Processors

“…In [18], the Harris operator is optimized using a number of optimizations such as vectorization, data interleaving and parallelization, on both x86/x64 and Arm processors. In [19], different ways of vectorizing the 3D convolution are shown. In [20], a white paper for a Gaussian Blur implementation on Intel processors is proposed, using FP computations.…”