DESIGN AND IMPLEMENTATION OF VLSI BASED HARDWARE ACCELERATORS FOR REAL TIME VIDEO IMAGE PROCESSING

Video image processing hardware implementation is continually driven to achieve
high performance e.g. high resolution, less power consumption, high throughput and
better image quality. However, the improvement has not achieved the required
performance to support demanding complex computation and large data transfer in
application such as video surveillance and telemedicine on portable devices which
require real time performance and with high data integrity. This work proposes low
memory bandwidth access within edge detection hardware accelerator design to
achieve real time performance of video image processing application in portable
device. An enhanced parallel and pipeline technique is utilized to perform transfer
data between memory and computation part and leading towards memory bandwidth
reduction to achieve real time performance. The hardware accelerator edge detection
design is integrated with clock divider, accelerator control, memory access controller,
memory access register, address counter, address decoder and acknowledge generator.
The design is compatible with standard interfacing of embedded processor system. It
includes the significant components in embedded processor system design such as
address decoder and address counter. These two components are involved in
controlling base address registers and address offset counters for the original and
derivative images, respectively. The hardware accelerator edge detection designs are
implemented on Altera Stratix III DSP development board and enables application of
co-processor without requiring new application specific digital signal processor. The
embedded video image processing with the integrated hardware accelerator edge
detection co-processor was integrated with Altera Quartus System-On- a�Programmable-Chip (SOPC). The implementation result shows a field programmable
gate arrays (FPGAs) acting as co-processor platforms for user defined co-processor,
with real time performance at a frame rate of 30 fps with a resolution of 720 x 480.
The parallel and pipeline technique are utilized in memory access, resulting more than
70% memory bandwidth reduction.