Researchers at the Lulea University of Technology (Lulea, Sweden) led by associate professor Matthew Thurley are involved in a project to improve the existing software in the Compute Unified Device Architecture (CUDA) image processing library by reducing the processing times of the morphological operators erosion and dilation.
CUDA was developed by NVIDIA to allow general parallel computation to be performed on CUDA enabled GPU graphics cards. Although NVIDIA's Performance Primitives (NPP) library contains fundamental morphological operators, the Swedish researchers want to improve their performance.
As such, they are developing freely available GPU morphological extensions (referred to as LTU-CUDA) to NVIDIA CUDA by determining where the current bottlenecks in performance are and where they might be improved.
Recently, Lulea University researcher Thurley and his student Victor Danell described an implementation of fast morphological erosion and dilation based on CUDA/NPP using the vHGW algorithm -- one of the fastest for computing grayscale image dilations and erosions on a serial CPU. The researchers say that their vHGW algorithm showed significant performance improvements over the generic CUDA library NPP.
A paper describing their work -- "Fast Morphological Image Processing Open-Source Extensions for GPU processing with CUDA" -- can be found here.
The LTU-CUDA is an ongoing project and the code is freely available here.
Editor's note: Associate professor Matthew Thurley is founder of a group in industrial image analysis at Lulea University of Technology where he has developed a series of industrial machine vision prototypes for the Swedish mining, aggregates and steel industries.
Related articles from Vision Systems Design you might also be interested in reading.
1. Motion estimation algorithm ported to GPU
Researchers from the Illinois Institute of Technology (Chicago, IL, USA) have taken a general purpose block-matching algorithm which is commonly used for motion estimation and ported it to run on multiple NVIDIA (Santa Clara, CA, USA) GPU cards using the Compute Unified Device Architecture (CUDA) computing engine.
2. Vision could benefit from parallel computing standard
In an effort to make it easier for programmers to take advantage of parallel processing hardware, Nvidia, Cray, the Portland Group (PGI), and CAPS enterprise have developed a new parallel-programming standard known as OpenACC.
3. MIT researchers develop new programming language for image processing
Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL; Cambridge, MA, USA) aim to make writing image-processing algorithms easier with a new programming language called Halide.
-- Dave Wilson, Senior Editor, Vision Systems Design
