Signal Compression for C3 Applications Using Hyperdistributions
Abstract
We develop our previous work on hyper-distributions into a formally well-defined transform which may be applied to images, the hyperdistribution transform (HDT). The HDT has many properties in common with conventional orthogonal transforms of signals, such as the Fast Fourier Transform, which suggests the possibility of developing a fast algorithm for the HDT. Presently, we have formulated the HDT in matrix language, which permits a reasonably efficient computational approach to calculating the HDT of an image. We then apply the HDT to image compression by representing the image as a truncated HDT expansion and reconstructing the image from the truncated HDT expansion. The compression ratio is measured in terms of the number of bits in the truncated HDT expansion compared to the number of bits in the original image. Test cases involving both synthetic and natural images are considered. Good quality reconstructions of natural images are obtained with compression ratios of 4:1 and recognizable images are obtained with compression ratios of 16:1. It was not necessary to segment the images into sub-images. Substantial further improvements in the performance of HDT compression may be obtained by employing image segmentation and other standard techniques for transform-based image compression algorithms.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 15, 1991
- Accession Number
- ADA270138
Entities
People
- Ta-ming Fang