Abstract
In digital image processing, the noise detection and removal are very important tasks, since they have wide applications in all fields. In recent years, the logic fabric and routing FPGAs architecture provides customers with a number of advantages. In this chapter, the performance is analysed with different FPGA processors in terms of slices, LUTs, and BRAM utilization is studied. The implemented hardware architecture with digital images plays an important role in all daily life applications, industrial applications such as image recognition, computer tomography, satellite television, space imaging, magnetic resonance imaging, etc., and also in areas of research and technology of geographical information systems and astronomy.
Top1. Introduction
Multimedia implementation on an embedded environment is called as embedded multimedia processing and applications. The embedded multimedia processing is categorized in two types, one is software based embedded system and another one is hardware based embedded system. Among the two types, hard ware embedded systems are faster and mostly implemented on general purpose processors like CPU or Reconfigurable Architectures like FPGAs. On the other hand, the same speed and flexibility can be achieved using software-based system with higher cost and increased complexity. In the design of computing architectures, the two factors must be considered and they are physical area and computing speed. These two factors affect the power consumptions as well as heat dissipation quantity. In recent years, high performance computing can be achieved by parallel processing, pipelining and multicore architectures. FPGA architectures can handle the parallel process effectively by configuring group of logic elements for each multimedia applications.
Recent FPGA architectures come with larger memory support and in-built processor so that high performance computing is much easier than CPUs because they have designed originally for specific set of sequences, difficult to custom for other applications. The power consumption of this CPUs also very high and not suitable for multimedia devices like mobile, camera, video recorders, etc. Other type of processor called Single Instruction Multiple Data (SIMD) supports parallel processing on hardware realization of multimedia due its larger cache memory size, but its performance depends on programming skills.
The multimedia content may be the image or video or audio or any other special format. In our works we consider only imaging applications on FPGA-Based embedded hardware. Our aim is to familiarize the state of art in compiling high level programs such as MatLab, C/C++, Open-CV, Python as well as Hardware Description Languages to FPGAs and to survey the relevant programming techniques and frameworks in FPGAs. This paper gives the consolidated works of various hardware realizations of image processing, imaging using CPUs/ Graphics Processing Units (GPUs), different FPGA architectures, various algorithms to achieve high performance and lower physical dimensions thus reduce heat dissipations.