Content-type: text/html Man page of vdpcmp

vdpcmp

Section: User Commands (1)
Index Return to Main Contents
 

NAME

vdpcmp - Compare two images using Visual Difference Predicator (low level command)  

SYNOPSIS

vdpcmp <target_file> <mask_file> [--ldr] [--mask <val>] [--psycho <val>] [--detection-mechanism <dmID>] [--no-phase-uncertainty] [--dump <pattern>] [--verbose] [--help]  

DESCRIPTION

This command processes a pair of images using Visual Difference Predicator and writes a map of probability detection to standard output. It is more convenient to use 'vdp' instead of this command, as it can handle image format conversion.

<target_file> is the file name of an image for which detection probability should be computed (e.g. distorted image) and <mask_file> is the file name of an image that is a reference (e.g. original, not distorted). The order in which both images are passed to the vdp is important and may affect the results.

The format of input images and output map is pfs. Use one of the readers from pfstools package to read images into vdpcmp.

 

OPTIONS

--ldr

Use the original (non high dynamic range) version of the VDP. Since the original VDP has not been maintained for ages, this option is depreciated. HDR VDP should provide better results and can work with LDR images as well.

--multiply-lum <val>

Multiply luminance values of both target and mask by <val>. Using such multiplier is necessary if the images contain only relative but not absolute luminance values (cd/m^2). In case of low dynamic range images that have been converted from sRGB color space to XYZ (--linear switch of pfsinppm), --multiply-lum should be set to the maximum luminance of the display (see also man page of vdp).

--mask <val>
Set the slope of the masking function to <val>. Default 1.

--psycho <val>
Set the slope of the psychometric function to <val>. Default 3.5.

--peak-contrast <contrast>
The peak sensitivity of the visual system given as minimum distinguishable contrast. The value must be given as a fraction, not percent!. Default value: 0.006 (=0.6%)

--otf <otfID>
Name of the optical tranfer function (OTF, or PSF of the eye). Possible values:

DEELEY (default) - OTF from 'Deeley, R.J., Drasdo, N., & Charman, W. N. (1991). A simple parametric model of the human ocular modulation transfer function. Ophthalmology and Physiological Optics, 11, 91-93'. It takes into account pupil size, which depends on adaptation luminance. This is suggested OTF function for the VDP.

MARIMONT - OTF from 'Marimont D.H. and Wandell B.A., 1994. Matching colour images; the effects of axial chromatic aberration. Journal of the Optical Society of America A, 11 (no 12) 2113-3122'. It takes into account pupil size, which depends on adaptation luminance. (note: the integration performed in the current implementation is not accurate. This can lead to artifacts in the filtered image).

SPENCER - PSF from Spencer et al. SIGGRAPH'95 paper

NORMANN - PSF from Normann et al. IEEE Trans. on Systems, Man and Cybernetics '83 paper (note: this function can be wrong (typo in the paper?), therefore it is not recommended to use it).

WESTHEIMER - PSF from Westheimer "The eye as an optical instrument" in Handbook of Perception and Visual Performance. This function is also used in Sarnoff Visual Discrimination Model.

NONE - no OTF preprecessing

--detection-mechanism <dmID>
Name of the detection mechanism to use. Currently supported:

WATSON, VDP - Watson cortex transform (default)

SIMPLE - no multichannel processing, only a single channel

--no-abs-fix

VDP can predict visible differences for pixels that are identical because of inaccuracies of visual band filters. To avoid this, as a final step HDR VDP compares two input images pixel by pixel and if the corresponding pixels are almost identical (difference < 0.0001), it sets the prediction for that pixel to 0. Use --no-abs-fix to disable this behavior.

--dump <pattern>
Dump intermediate images resulting from VDP processing. The pattern specifies names of the files to dump. A shell pattern ('*' is a wild card character) can be used to dump many frames. The option can be specified many times to dump additional images. All images are dumped in pfs format. Some available dump images are listed below. More can be found in the sources.
target_otf.pfs
target after OTF filtering
mask_otf.pfs
mask after OTF filtering
filter_otf.pfs
OTF filter
filter_csf_N.pfs
CSF filter for the adaptation luminance 10^N
adaptation_map.pfs
adaptatin map used for CSF filtering
target_photoreceptor_nonlin.pfs
target after photoreceptor nonlinearity
mask_photoreceptor_nonlin.pfs
mask after photoreceptor nonlinearity
target_csf.pfs
target after CSF filtering
mask_csf.pfs
mask after CSF filtering
dom_filter_cortex_N.pfs
DOM filter for the band N (1-6)
mask_cortex_N_M.pfs
mask for the band N and orientatiom M after cortex transform
target_cortex_N_M.pfs
target for the band N and orientatiom M after cortex transform
filter_cortex_N_M.pfs
cortex filter for the band N and orientatiom M
difference_cortex_N_M.pfs
difference between target and mask for the band N and orientatiom M
te_cortex_N_M.pfs
threshold elevation (due to masking) for the band N and orientatiom M
probability_cortex_N_M.pfs
probability of detection for the band N and orientatiom M
--no-phase-uncertainty
Phase uncertainty improves the prediction of visual masking by discounting the influence of phase from the masking thresholds. By default, phase uncertainty is included in the VDP calculation. However it almost doubles time required to process an image. This switch can be used to turn off computation of phase uncertainty.

The following options define viewing conditions. Because passing a complete set of options to specify viewing conditions is rather tedious, the better way is to use -vc option of the vdp command (see the man page for vdp).

--display-width --display-height
Physical dimensions of the display given in meters. Default: width 0.375 height 0.300.

--min-distance --max-distance
Minimum and maximum distance from which the display is observed. Default: min 0.5 max 0.5.

--display-x-resolution --display-y-resolution
Vertical and horizontal resolution of the display given in pixels. Default: horizontal 1280 vertical 1024.

 

SEE ALSO

vdp(1) vdpvis(1) pfsin(1)  

BUGS

Please report bugs and comments to Rafal Mantiuk <mantiuk@mpi-sb.mpg.de>.


 

Index

NAME
SYNOPSIS
DESCRIPTION
OPTIONS
SEE ALSO
BUGS

This document was created by man2html, using the manual pages.
Time: 16:48:18 GMT, July 02, 2008