Image compression using BinDCT for Dynamic Hardware FPGAs

ABSTRACT ____________________________________________________________ This thesis investigates the prospect of using a Binary Discrete Cosine Transform as an integral component of an image compression system. The Discrete Cosine Transform (DCT) algorithm is well known and commonly used for image compression. Various compression techniques are actively being researched as they are attractive for many industrial applications. The particular compression technique focused on was still image compression using the DCT. The recent expansion of image compression algorithms and multimedia based mobile, including many wireless communication applications, handheld devices, digital cameras, videophones, and PDAs has furthered the need for more efficient ways to compress both digital signals and images.___________________________________________________________ This thesis investigates the prospect of using a Binary Discrete Cosine Transform as an integral component of an image compression system. The Discrete Cosine Transform (DCT) algorithm is well known and commonly used for image compression. Various compression techniques are actively being researched as they are attractive for many industrial applications. The particular compression technique focused on was still image compression using the DCT. The recent expansion of image compression algorithms and multimedia based mobile, including many wireless communication applications, handheld devices, digital cameras, videophones, and PDAs has furthered the need for more efficient ways to compress both digital signals and images. The objective of this research to find a generic model to be used for image compression was met. This software model uses the BinDCT algorithm and also develops a detection system that is accurate and efficient for implementation in hardware, particularly to run in real-time. This model once loaded on to any dynamic hardware should update by reconfiguring the FPGA automatically, during run time with different BinDCT processors. Such a model will enhance our understanding of the dynamic BinDCT processor in image compression. Image analysis involves examination of the image data for a specific application. The characteristic of an image decides the most efficient algorithm. Selection techniques were designed centred on use of the entropy calculation for each 8 x 8 tile. However many other techniques were analysed such as homogeneity. Selection of the most efficient BinDCT algorithm for each tile was a challenge met by analysis of the entropy data. For the BinDCT different configurations were analysed with standard grey scale photographic images. Upgrading the available technology to the point where the most suitable BinDCT configuration for each image tile input stream will be continuously configured all the time, will lead to significant coding advantage in image analysis and traditional