Efficient Image Re-Ranking Computation on GPUs

Abstract

The huge growth of image collections and multimedia resources available is remarkable. One of the most common approaches to support image searches relies on the use of Content-Based Image Retrieval (CBIR) systems. CBIR systems aim at retrieving the most similar images in a collection, given a query image. Since the effectiveness of those systems is very dependent on the accuracy of ranking approaches, reranking algorithms have been proposed to exploit contextual information and improve the effectiveness of CBIR systems. Image re-ranking algorithms typically consider the relationship among every image in a given dataset when computing the new ranking. This approach demands a huge amount of computational power, which may render it prohibitive on very large data sets. In order to mitigate this problem, we propose using the computational power of Graphics Processing Units (GPU) to speedup the computation of image re-ranking algorithms. GPUs are fast emerging and relatively inexpensive parallel processors that are becoming available on a wide range of computer systems. In this paper, we propose a parallel implementation of an image re-ranking algorithm designed to fit the computational model of GPUs. Experimental results demonstrate that relevant performance gains can be obtained by our approach.

Contextual Spaces Re‐Ranking: accelerating the Re‐sort Ranked Lists step on heterogeneous systems

“…Two alternatives have recently emerged as possible solutions: rank-based methods [5,17,18], in which the information of top-k retrieved images is exploited, reducing the required computation, and the use of parallel and heterogeneous architectures [7,19,20], made popular for accelerating several image applications.…”

“…Parallel solutions and heterogeneous computing were exploited for accelerating some methods [7,8], but open research challenges still exist. For example, despite the diversity of the approaches, these methods [2,5,7,8] present a common characteristic related to the use of iterative sorting steps, which can be computationally expensive for large datasets.…”

“…

Re-ranking algorithms have been proposed to improve the effectiveness of content-based image retrieval systems by exploiting contextual information encoded in distance measures and ranked lists. In a real-world scenario, CBIR systems require both good effectiveness and efficiency (response time), so re-ranking methods must be improved.Central processing units (CPUs) no longer have just one core, and graphics processing units (GPUs) are now being used as general purpose processors due to having evolved into massive parallel architectures capable of executing hundreds of operations per cycle [7]. One of our approaches consists in parallelizing the algorithm with OpenCL to use the central and graphics processing units of an accelerated processing unit.

…”

Effective, Efficient, and Scalable Unsupervised Distance Learning in Image Retrieval Tasks

“…A possible solution is to ensure exclusive access, but the overhead associated with fine grained synchronization in OpenCL is significant. Therefore, as suggested by other works [30], we allow direct updates to matrix A because the loss of some updates has a very low impact on the effectiveness of the algorithm (as demonstrated by confidence intervals of Section 4.4).…”

“…Since most of changes occur only in the beginning of the ranked lists [30], we use the insertion sort algorithm, whose complexity tends to be linear when the input is almost sorted.…”