More Info

The application of digital image processing is today extremely diversified: multimedia, digital television, satellite imagery, medical or military imagery, industrial vision, etc. These domains lead to many applications such as image database management, compression of still or moving images for storage or transmission, analysis and recognition of objects, etc … Any manipulation of images puts at stake a considerable volume of data, this one is related to the dimension of the images manipulated by application and increases when, to provide an element of interpretation, it is to have access to several consecutive images of a sequence. 

A conventional image processing chain comprises several processing levels of the image acquisition elements, then steps of filtering, extraction of parameters (contours, shapes, etc.) and, depending on the applications, steps of Going from the beginning of the chain to its end, the amount of processed data decreases, resulting in a significant decrease in the computing power required. At the same time, the complexity of algorithms is increasing. Requiring the user to modify this for each image manipulation.

Any image manipulation involves a considerable amount of data, so the image processing computation volume often exceeds the processing capacity of a conventional machine. The problem is even more pronounced in a real-time context (follow the video rate for example). Fortunately, several parallelization techniques can be applied to image processing algorithms.

The most commonly used approach is to distribute the image between the processors by storing each pixel of a sub-image in the same memory module. The image is thus divided into sub-images.

Speed can be increased by distributing individual processes (image processing algorithm steps) across multiple cores of the processor. However, this requires rewriting the algorithms in a form that allows this distribution.

RabbitFire^tm uses a parallel pipeline approach that allows multiple instances (images) to be processed by multiple processors, taking into account the needs and / or restrictions of each of them.This approach offers the advantage of minimizing the burden and complication of the code. This however adds to the complexity due to the need to manage a variety of strategies, data synchronization and temporary storage.

RabbitFire^tm provides a simple and effective solution that addresses these complexities by significantly simplifying the process of building a parallel application.

RabbitFire^tm is a parallel processing software that makes it easy for developers and non-developers to create or enhance applications that run on multiple cores / processors and dramatically improve performance.

RabbitFire^tm makes it possible to understand and optimize image processing thanks to a visual monitoring tool.

How RabbitFire^tm Simplifies Parallel Processing

RabbitFire^tm enables developers to run applications that process stream data, such as audio and video, on many processors simultaneously to dramatically boost performance.