Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: GPU algorithm

Sortuj według:

Ogranicz wyniki do:

A New Idea of Fast Three-Dimensional Median Filtering for Despeckling of Optical Coherence Tomography Images

Gocławski J., Sekulska-Nalewajko J.

Image Processing & Communications

2015

Vol. 20, no. 3

25--34

Median filtering has been widely used in image processing for noise removal because it can significantly reduce the power of noise while limiting edge blurring. This filtering is still a challenging task in the case of three-dimensional images containing up to a billion of voxels, especially for large size filtering windows. The authors encountered the problem when applying median filter to speckle noise reduction in optical coherence tomography images acquired by the Spark OCT systems. In the paper a new approach to the GPU (Graphics Processing Unit) based median smoothing has been proposed, which uses two-step evaluation of local intensity histograms stored in the shared memory of a graphic device. The solution is able to output about 50 million voxels per second while processing the neighbourhood of 125 voxels by Quadro K6000 graphic card configured on the Kepler architecture.

Real time area-based stereo matching algorithm for multimedia video devices

Hachaj T., Ogiela M. R.

Opto-Electronics Review

2013

Vol. 21, No. 4

367-375

In this paper we investigate stereovision algorithms that are suitable for multimedia video devices. The main novel contribution of this article is detailed analysis of modern graphical processing unit (GPU)-based dense local stereovision matching algorithm for real time multimedia applications. We considered two GPU-based implementations and one CPU implementation (as the baseline). The results (in terms of frame per second, fps) were measured twenty times per algorithm configuration and, then averaged (the standard deviation was below 5%). The disparity range was [0,20], [0,40], [0,60], [0,80], [0,100] and [0,120]. We also have used three different matching window sizes (3×3, 5×5 and 7×7) and three stereo pair image resolutions 320×240, 640×480 and 1024×768. We developed our algorithm under assumption that it should process data with the same speed as it arrives from captures’ devices. Because most popular of the shelf video cameras (multimedia video devices) capture data with the frequency of 30Hz, this frequency was threshold to consider implementation of our algorithm to be “real time”. We have proved that our GPU algorithm that uses only global memory can be used successfully in that kind of tasks. It is very important because that kind of implementation is more hardware-independent than algorithms that operate on shared memory. Knowing that we might avoid the algorithms failure while moving the multimedia application between machines operating different hardware. From our knowledge this type of research has not been yet reported.