Visual Communication - Information Rate and Visual Quality

Information Rate and Visual Quality

 
 
 
 
 
 
The two targets shown here are used as examples. Images of the resolution wedges, with expanding width times the sampling intervals, are highly sensitive to the perceptual defects due to blurring and aliasing. The random polygons represent a particular realization of scenes with the PSD used to compute the information rate as a funtion of the SFR and SNR of the image-gathering device and the mean spatial detail relative to the sampling intervals. The figures below show the corresponding images restored with the Wiener filter for two of the informationally-optimized designs summarized in the Table.

The above images r(x,y) consist of three components

r(x,y)=(x,y) * r(x,y)+nar(x,y) + npr(x,y)

where

(x,y) * r(x,y) accounts for the blurring by the throughput response r(x,y) (see result Throughput response),

nar(x,y) accounts for the aliasing caused by the insufficient sampling that occurs because the SFR extends beyond the sampling passband , and npr(x,y) accounts for the photodetector noise.

The extraneous periodic structure that aliasing produces in the images of the resolution wedges is commonly referred to as Moire pattern. The corresponding distortion in the images of the random polygons, however, emerges more subtle as jagged (or staircase) edges. If these images are displayed at a small format, then the jagged edges could not be resolved by the observer and would appear to be the result of blurring instead of aliasing. Aliasing has therefore often been overlooked as a significant source of image degradation.

The images produced by Wiener restorations, such as the ones shown above, normally have a high resolution and sharpness. However, it is often desirable to combine this restoration with an enhancement filter that gives the user some control over the trade-off among fidelity, resolution, sharpness and clarity. The images shown below represent such enhancement for the above images.

Place cursor over image to zoom in

 
 
The images that the observer sees on the monitor as a halftone print is subject to some loss in information due to the limited number of distinguishable gray levels of the image-display medium. As shown here, this loss may be small for designs 2 and 3 but substantial for design 1. Hence, the images for design 1 cannot convey fully the improvement in image quality that higher quality image-display media can produce.


 
 

Results

Robustness

Home Site map


 
 
Web site curator: Glenn Woodell