Spatial Subsampling Research Articles

Holographic bandwidth compression using spatial subsampling

Subject terms: electro-holography, holographic computation, holographic bandwidth compression, diffraction- specific, fringelet, vector Abstract. A novel electro-holographic bandwidth compression technique, fringelet bandwidth compression, is described and implemented. This technique uses spatial subsampling to reduce the bandwidth and complex- ity of holographic fringe computation for real-time 3-D holographic displays. Fringelet bandwidth compres- sion is a type of diffraction-specific fringe computation, an approach that considers only the reconstruction process in holographic imaging. The fringe pattern is treated as a spectrum that is sampled in space (as holo- graphic elements or hogels) and in spatial frequency (as hogel vectors). Fringelet bandwidth compression achieves a compression ratio of 16:1 without conspicuously degrading image quality. Further increase in com- pression ratio and speed is possible with additional image degradation. Fringelet decoding is extremely simple, involving the replication of fringelet sample values. This simplicity enables an overall increase in fringe com- putation speed of over 3000 times compared to conventional interference-based methods. The speed of fringe- let bandwidth compression has enabled the generation of images at nearly interactive rates: under 4.0 s per hand-sized (one-liter) 3-D image generated from a 36-Mbyte fringe.

Low bit-rate video compression with neural networks and temporal subsampling

In this paper we describe a novel neural network technique for video compression, using a point-process type neural network model we have developed which is closer to biophysical reality and is mathematically much more tractable than standard models. Our algorithm uses an adaptive approach based upon the users' desired video quality Q, and achieves compression ratios of up to 500:1 for moving gray-scale images, based on a combination of motion detection, compression, and temporal subsampling of frames. This leads to a compression ratio of over 1000:1 for full-color video sequences with the addition of the standard 4:1:1 spatial subsampling ratios in the chrominance images. The signal-to-noise ratio ranges from 29 dB to over 34 dB. Compression is performed using a combination of motion detection, neural networks, and temporal subsampling of frames. A set of neural networks is used to adaptively select the desired compression of each picture block as a function of the reconstruction quality. The motion detection process separates out regions of the frame which need to be retransmitted. Temporal subsampling of frames, along with reconstruction techniques, lead to the high compression ratios.

Relating seasonal patterns of the AVHRR vegetation index to simulated photosynthesis and transpiration of forests in different climates

Recent research has suggested that the Normalized Difference Vegetation Index (NDVI) calculated from the AVHRR sensor is directly related to photosynthesis (PSN), transpiration (TRAN), and aboveground net primary production (ANPP) of terrestrial vegetation. Weekly NDVI data for 1983–1984 were retrieved for seven sites of diverse climate in North America. The sites were Fairbanks, AK, Seattle, WA, Missoula, MT, Madison, WI, Knoxville, TN, Jacksonville, FL, and Tucson, AZ. Meteorological data from ground stations were retrieved to drive an ecosystem simulation model (FOREST-BGC) calculating daily canopy PSN and TRAN and annual ANPP of a hypothetical forest stand for the corresponding period at each site. Correlations of annual integrated NDVI across all sites for both years were: annual PSN, R 2 = 0.87; annual TRAN, R 2 = 0.77; annual ANPP, R 2 = 0.72. Correlation between weekly NDVI and PSN was variable; with high latitude wet sites, R 2 = 0.77–0.84. On sites with less seasonal amplitude of NDVI and PSN, or on sites with substantial seasonal water stress correlations ranged from R 2 = 0.08 to 0.64. Correlations of weekly NDVI with TRAN followed the same pattern as PSN, but were slightly lower. The tendency of raw NDVI data to overpredict PSN and TRAN on water limited sites was partially corrected using an “aridity index” of annual radiation/annual precipitation that could be computed from general climatological data for improving large scale NDVI maps of PSN and TRAN. The spatial subsampling done for the global vegetation index (GVI) precludes following specific study sites through the growing season. We conclude that estimates of vegetation productivity using the GVI should only be done as annual integrations until unsubsampled local area coverage (LAC) NDVI data can be tested against forest PSN, TRAN, and ANPP, measured at shorter time intervals.

Some new formulae applicable to electrochemical nucleation/growth/collision

This paper presents some new formulae applicable to electrochemical nucleation/growth/collision, and is divided into six parts. In the first part general formulae for probabilistic effects in space, time and energy are presented. Then in the second part a new paradigm for 2-D nucleation/growth/collision at constant potential is described. Part three summarizes our recent work on the diffusion-controlled growth of hemispherical nuclei, including the effects of non-equilibrium and preceding chemical reaction, while part four consists of a discussion of nucleation as a stochastic process, and includes formulae for spatial and temporal subsampling. Part five discusses the occurrence of inductive loops in sinusoidal and triangular potential scan measurements of nucleating systems, and finally part six lists some formulae in integral geometry applicable to crystal growth problems.

Spatial Subsampling in Motion-Compensated Television Coders

Motion-compensated television coders generate data at a nonuniform rate, which is smoothed by a buffer for transmission over a channel of constant bit rate. Spatial subsampling is one of the methods used to prevent overflow of the buffer, which would otherwise occur for scenes with complex motion from frame to frame. In this paper we evaluate the effects of spatial subsampling on the performance of motion-compensated coders. In particular, we find that, although the quality of motion estimation does degrade in the presence of subsampling, the degradation is not substantial. Use of 2:1 horizontal subsampling, for example, results in bit rates that are 50 percent lower compared with no subsampling for motion-compensated coders. This percentage is approximately the same for the conditional-replenishment coders. Spatial subsampling generally results in blurring of the picture. We describe a technique for adaptive interpolation that results in blurring of only the unpredictable areas of the picture. The subjective quality in the presence of subsampling is thus improved considerably. In conclusion, our techniques for subsampling in a motion-compensated coder reduce the bit rate approximately by the same factor as subsampling in a conditional-replenishment coder, but result in a much better picture quality.