A Real-Time Image Compressor Using 2-Dimensional DWT and Its FPGA Implementation

Young-Ho SEO  Wang-Hyun KIM  Ji-Sang YOO  Dai-Gyoung KIM  Dong-Wook KIM  

Publication
IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E87-A   No.8   pp.2110-2119
Publication Date: 2004/08/01
Online ISSN: 
DOI: 
Print ISSN: 0916-8508
Type of Manuscript: PAPER
Category: VLSI Design Technology and CAD
Keyword: 
DWT,  wavelet,  compression,  FPGA,  design,  hardware,  

Full Text: PDF>>
Buy this Article




Summary: 
This paper proposes the design and implementation of a real-time image compressor using 2-Dimensional Discrete Wavelet Transform (2DDWT), which targets an FPGA as its platform. The image compressor uses Daubechies' bi-orthogonal DWT filters (9, 7) and 16-bit fixed-point data formats for wavelet coefficients in the internal calculation. The target image is NTSC 640240 pixels per field whose color format is Y:Cb:Cr = 4:2:2. We developed for the 2DDWT a new structure with four Multipliers and Accumulators (MACs) for real-time operations. We designed and used a linear fixed scalar quantizer, which includes the exceptional treatment of the coefficients whose absolute values are larger than the quantization region. Only a Huffman entropy encoder was included due to the hardware overhead. The quantizer and Huffman encoder merged into a single functional module. Due to the insufficient memory space of an FPGA, we utilized external memory (SDRAM) as the working and memory storage space. The proposed image compressor maps into an APEX20KC EP20K600CB652-7 from Altera and uses 45% of the Logic Array Block (LAB) and 9% of the Embedded System Block (ESB). With a 33 MHz clock frequency, the proposed image compressor shows a speed of 67 fields per second (33 frames per second), which is more than real-time operation. The resulting image quality from reconstruction is approximately 28 dB in PSNR and its compression ratio is 29:1. Consequently, the proposed image compressor is expected to be used in a dedicated system requiring an image-processing unit.