Predicting human vision perception and perceptual difference

1. A perceptual difference model using a pair of human vision systems, one for a reference video signal and the other for an impaired version of the reference video signal, each human vision system having a spatio-temporal filter receiving the respective video signal, the outputs of which are differenced to obtain an initial perceptual difference map, further comprising means for noise masking the output of each spatio-temporal filter to account for photon noise based on the relation of pupil size to luminance prior to obtaining the initial perceptual difference map.

2. The perceptual difference model as recited in claim 1 further comprising means for masking the initial perceptual difference map as a function of the outputs of the noise masking means and filtered luminance outputs from the spatio-temporal filters to account for correlation and contrast gain control to produce an output perceptual difference map from the initial perceptual difference map.

3. The perceptual difference model as recited in claim 2 wherein the masking means comprises:

means for obtaining an absolute noise masked sum value and a noise masked product value from the outputs of the noise masking means;

means for generating a correlation value from the noise masked product value;

means for obtaining an absolute luminance sum value and a luminance product value from the filtered luminance outputs;

means for combining the absolute noise masked sum value, the correlation value, the absolute luminance sum value and the luminance product value to produce a contrast gain mask;

means for generating an area threshold mask from the absolute noise masked sum value and the absolute luminance sum value; and

means for adjusting the initial perceptual difference map with the area threshold mask and the contrast gain mask to produce an output perceptual difference map.

4. The perceptual difference model as recited in claim 2 , or 3 wherein each noise masking means comprises:

means for obtaining a photon noise amplitude estimation as a noise signal based on the relation of pupil size to luminance from the filtered luminance output of the corresponding spatio-temporal filter; and

means for generating a noise masked signal from the noise signal and the output of the spatio-temporal filter as a source signal, the generating means being a function of the square root of the sum of the squares of the noise signal and the source signal less the noise signal, the respective noise masked signals being differenced to obtain the initial perceptual difference map.

5. A perceptual difference model using a pair of human vision systems, one for a reference video signal and the other for an impaired version of the reference video signal, each human vision system having a spatio-temporal filter receiving the respective video signal, the outputs of which are differenced to obtain an initial perceptual difference map, further comprising means for masking the initial perceptual difference map as a function of the outputs of the spatio-temporal filters and filtered luminance outputs from the spatio-temporal filters to account for correlation and contrast gain control to produce an output perceptual difference map from the initial perceptual difference map.

6. The perceptual difference model as recited in claim 5 wherein the masking means comprises for each pixel at each time sample:

means for obtaining an absolute filtered sum value and a filtered product value from the outputs of the spatio-temporal filters;

means for generating a correlation value from the filtered product value;

means for obtaining an absolute luminance sum value and a luminance product value from the filtered luminance outputs;

means for combining the absolute filtered sum value, the correlation value, the absolute luminance sum value and the luminance product value to produce a contrast gain mask;

means for generating an area threshold mask from the absolute filtered sum value and the absolute luminance sum value; and

means for adjusting the initial perceptual difference map with the area threshold mask and the contrast gain mask to produce an output perceptual difference map.

7. A method of generating an output perceptual difference map from a reference video signal and an impaired version of the reference video signal using a perceptual difference model that merges a pair of human vision systems, one for the reference video signal and the other for the impaired video signal, each human vision system having a spatio-temporal filter receiving the respective video signal, the outputs of which are differenced to obtain an initial perceptual difference map, further comprising the step of noise masking the output of each spatio-temporal filter to account for photon noise based on the relation of pupil size to luminance prior to obtaining the initial perceptual difference map as the output perceptual difference map.

8. The method as recited in claim 7 further comprising the step of masking the initial perceptual difference map as a function of the outputs of the noise masking step and filtered luminance outputs from the spatio-temporal filters to account for correlation and contrast gain control to produce the output perceptual difference map from the initial perceptual difference map.

9. The method as recited in claim 8 wherein the masking step comprises the steps of:

obtaining an absolute noise masked sum value and a noise masked product value from the outputs of the noise masking step;

generating a correlation value from the noise masked product value;

obtaining an absolute luminance sum value and a luminance product value from the filtered luminance outputs;

combining the absolute noise masked sum value, the correlation value, the absolute luminance sum value and the luminance product value to produce a contrast gain mask;

generating an area threshold mask from the absolute noise masked sum value and the absolute luminance sum value; and

adjusting the initial perceptual difference map with the area threshold mask and the contrast gain mask to produce the output perceptual difference map.

10. The method as recited in claim 7 , 8 or 9 wherein each noise masking step comprises the steps of:

obtaining a photon noise amplitude estimation as a noise signal based on the relation of pupil size to luminance from the filtered luminance output of the corresponding spatio-temporal filter; and

generating a noise masked signal from the noise signal and the output of the spatio-temporal filter as a source signal, the generating means being a function of the square root of the sum of the squares of the noise signal and the source signal less the noise signal, the respective noise masked signals being differenced to obtain the initial perceptual difference map.

11. A method of generating an output perceptual difference map from a reference video signal and an impaired version of the reference video signal using a perceptual difference model that merges a pair of human vision systems, one for the reference video signal and the other for the impaired video signal, each human vision system having a spatio-temporal filter receiving the respective video signal, the outputs of which are differenced to obtain an initial perceptual difference map, further comprising the step of masking the initial perceptual difference map as a function of the outputs of the spatio-temporal filters and filtered luminance outputs from the spatio-temporal filters to account for correlation and contrast gain control to produce the output perceptual difference map from the initial perceptual difference map.

12. The method as recited in claim 11 wherein the masking step comprises the steps of:

obtaining an absolute filtered sum value and a filtered product value from the outputs of the spatio-temporal filters;

generating a correlation value from the filtered product value;

obtaining an absolute luminance sum value and a luminance product value from the filtered luminance outputs;

combining the absolute filtered sum value, the correlation value, the absolute luminance sum value and the luminance product value to produce a contrast gain mask;

generating an area threshold mask from the absolute filtered sum value and the absolute luminance sum value; and

adjusting the initial perceptual difference map with the area threshold mask and the contrast gain mask to produce the output perceptual difference map.

13. The method as recited in claim 8 , 9 , 11 or 12 further comprising the step of calibrating the perceptual difference model in response to known reference and impaired video signal stimuli by adjusting various parameters of the otherwise masking step so that the output perceptual difference map approximately matches known results for the known reference and impaired video signal stimuli.