S.M. Hanif and K. Abe (Japan)
Wavelet, sequence length conversion, and partial retrieval
Wavelet transformation has been adopted in the recent JPEG2000 image compression standard considering its multifarious advantages. But due to its inherent computational complexity, processing the whole image at one time in a mobile port will result in a higher circuit and memory requirements. If the image can be transmitted to the end port in a way so that it can be decoded part by part, these requirements can be lowered. It can be achieved by dividing the transformed coefficient array to smaller length sequences in an intermediate gateway, while sending an image to a mobile port from another system. An effective algorithm for dividing the image coefficient array through retrieving pixels near the division border and then correcting the coefficients for symmetric extension has been proposed earlier. In this paper we will investigate the approach of sequence length conversion through partial retrieval of the intermediate coefficients rather than retrieving the pixels themselves. We will show that through this approach it is possible to increase the processing and memory efficiency in the gateway processor, while containing the noise below a permissible limit. Improvement in PSNR through overflow correction in post-processing is also discussed.
Important Links:
Go Back
