Accelerating Translational Image Registration for HDR Images on GPU

TL;DR

This paper presents a GPU-accelerated method for fast and robust alignment of HDR images captured with hand-held cameras, significantly improving processing speed over traditional CPU methods.

Contribution

The study introduces a parallel GPU implementation of median threshold bitmap-based image registration, achieving substantial speed-up for HDR image alignment.

Findings

Achieves up to 6.24x speed-up over CPU implementation

Effective parallelization of MTB-based registration on GPU

Source code publicly available for reproducibility

Abstract

High Dynamic Range (HDR) images are generated using multiple exposures of a scene. When a hand-held camera is used to capture a static scene, these images need to be aligned by globally shifting each image in both dimensions. For a fast and robust alignment, the shift amount is commonly calculated using Median Threshold Bitmaps (MTB) and creating an image pyramid. In this study, we optimize these computations using a parallel processing approach utilizing GPU. Experimental evaluation shows that the proposed implementation achieves a speed-up of up to 6.24 times over the baseline multi-threaded CPU implementation on the alignment of one image pair. The source code is available at https://github.com/kadircenk/WardMTBCuda