The advancements in the field of internet and cloud computing has resulted in a huge amount of multimedia data and processing of this data have become more complex and computationally intensive. As a result, it has become very challenging for image retrieval algorithms to efficiently extract useful information from these data. Local Derivative Pattern (LDP) is a higher order local pattern operator used for image retrieval. Originally proposed for face recognition, LDP encodes the distinctive spatial relationships contained in a local region of an image as the feature vector. However LDP results in a very large feature vector thereby becoming computationally very expensive. In this paper, we propose efficient techniques for extracting parallelism from LDP algorithm and propose strategies for implementing it on GPGPUs. We show that with the optimal configuration of GPGPU kernels we can perform image retrieval at a much faster rate that would facilitate improved performance for image retrieval applications. We show that by porting LDP algorithm on CUDA a speedup of the order of 36X can be achieved as compared to its sequential counterpart for images having resolution of 192x192, 256x256 and 512x512 pixels.

<p><span>Pattern based texture descriptors are widely used in Content Based Image Retrieval (CBIR) for efficient retrieval of matching images. Local Derivative Pattern (LDP), a higher order local pattern operator, originally proposed for face recognition, encodes the distinctive spatial relationships contained in a local region of an image as the feature vector. LDP efficiently extracts finer details and provides efficient retrieval however, it was proposed for images of limited resolution. Over the period of time the development in the digital image sensors had paid way for capturing images at a very high resolution. LDP algorithm though very efficient in content-based image retrieval did not scale well when capturing features from such high-resolution images as it becomes computationally very expensive. This paper proposes how to efficiently extract parallelism from the LDP algorithm and strategies for optimally implementing it by exploiting some inherent General-Purpose Graphics Processing Unit (GPGPU) characteristics. By optimally configuring the GPGPU kernels, image retrieval was performed at a much faster rate. The LDP algorithm was ported on to Compute Unified Device Architecture (CUDA) supported GPGPU and a maximum speed up of around 240x was achieved as compared to its sequential counterpart.</span></p>

No abstract

No abstract

No abstract