Multi-core Image Processing: Q&A from DALSA

By Inder Kohli, DALSA Corporation, Montreal Edited by Marc Fimeri Adept Electronic Solutions

What is multi-core image processing?
Advances in silicon integration has allowed processor vendors to integrate more than one processing core on the same silicon die. Several processor vendors such as Intel, AMD, NVIDIA, AMD, TI etc. are providing multicore processors. The multiplicity of processing elements permits using "divide and conquer" algorithms to achieve higher processing throughput without necessarily having to increase clock rates. Higher clock rates also imply higher heat dissipation and related challenges. Keeping this in mind, "multicore image processing" can be defined as any image processing algorithm that exploits the multicore characteristic of a processor.

Q: How is multicore image processing implemented?
A: Multicore image processing is a technique by which the execution of an image processing application is distributed over different cores of a CPU.There are basically two methods for partitioning processing tasks on multiple CPU cores. The first one (frame-based partitioning) implies that a given core sequentially executes the whole algorithm on a given image frame. This sort of partitionning is implemented using multi-threading at the application level. The second one (pixel-based partitionning) involves splitting an image into sections and assigning each section to a given core. This method is implemented at the library level and is therefore transparent to the application. The choice between the two techniques often depends on the kind of image processing algorithm required by the application.

Q: What benefits does multicore image processing bring to the user?
A: Multicore image processing offers higher processing speeds at lower cost and at lower operating temperatures. Greater advantage can be taken from the latest PC processor architectures: can run faster and do more. It can minimise response time. More object-inspections per minute. More performance analysis operations possible from the application. Achieve more challenging processing tasks that are currently too time-consumptive.

Q: What challenges might a user face when moving from a single to multicore system?
A: Most of the image processing libraries and applications on the market were not designed with multiple cores in mind. Moreover, not all image processing tasks are well suited for partitioning on multicore processors and thus performance doesn't' scale with the available number of processing cores. The image processing tasks that involve large amounts of data movement also achieve limited improvements in processing speeds. Processor vendors employ different micro-architectures to implement multiple cores on the same die. As a result, multicore imaging libraries are developed and optimized for specific processor architectures. This limits the portability of the libraries across different processor architectures.

Q: What application areas are there for multicore image processing within the imaging and machine vision industry?
A: Virtually all areas of the imaging and machine vision industry will benefit from multicore processing. However, the extent to which benefits can be derived depends on the specific function being used.

Q: How do you see multicore image processing developing over the next few years?
A: In general, there is a high degree of interest in multicore processors and image processing in particular. Unified instruction sets and multicore aware operating systems will automatically scale applications to take full advantage of the underlying hardware.

Q: How does DALSA use multicore image processing? What products do you have within multicore image processing? Will you be developing more multicore image processing products in the future?
A: DALSA considers multicore optimized imaging libraries to be an important part of its software product roadmap. Earlier this year, DALSA launched a multi-core optimised version of its flagship imaging library, Sapera Essential.

Who benefits from using multi-core image processing?
Currently, virtually any industry using imaging and machine vision technologies can benefit from implementing multi-core processing. Future demands for this type of technology will intensify due to increasingly complex algorithms, faster cameras and improved resolutions.