US 5,940,530 A · Backlit scene and people scene detecting method and apparatus and a gradation correction apparatus

A backlit scene detecting method for detecting a backlit scene in an input image comprising the steps of:

dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
detecting a luminance of each block;
detecting dark blocks among the plurality of blocks, each dark block having a detected luminance which is smaller than a predetermined luminance;
counting a number of first continuous dark blocks in each row which extend from a right side edge of the input image and obtaining a first sum (S0) of the number of first continuous dark blocks from all the rows;
counting a number of second continuous dark blocks in each row which extend from a left side edge of the input image and obtaining a second sum (S1) of the number of second continuous dark blocks from all the rows;
counting a number of third continuous dark blocks in each column which extend from a top side edge of the input image and obtaining a third sum (S2) of the number of third continuous dark blocks from all the columns;
counting a number of fourth continuous dark blocks in each column which extend from a bottom side edge of the input image and obtaining a fourth sum (S3) of the number of fourth continuous dark blocks from all the columns;
and
determining that the input image is a backlit scene when either one of the following two conditions is satisfied:
(i) when the fourth sum (S3) is the greatest of the first to fourth sums;
and
(ii) when the third sum (S2) is the greatest and the fourth sum (S3) is the second greatest of the first to fourth sums.

9 more elements

A backlit scene detector for detecting a backlit scene in an input image comprising:

dividing means for dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
luminance detecting means for detecting a luminance of each block;
dark block detecting means for detecting dark blocks among the plurality of blocks, each dark block having a detected luminance which is smaller than a predetermined luminance;
dark block processing means for counting a number of first continuous dark blocks in each row which extend from a right side edge of the input image and obtaining a first sum (S0) of the number of first continuous dark blocks from all the rows, for counting a number of second continuous dark blocks in each row which extend from a left side edge of the input image and obtaining a second sum (S1) of the number of second continuous dark blocks from all the rows, for counting a number of third continuous dark blocks in each column which extend from a top side edge of the input image and obtaining a third sum (S2) of the number of third continuous dark blocks from all the columns, and for counting a number of fourth continuous dark blocks in each column which extend from a bottom side edge of the input image and obtaining a fourth sum (S3) of the number of fourth continuous dark blocks from all the rows;
and
backlit scene detecting means for detecting that the input image is a backlit scene when either one of the following two conditions is satisfied:
(i) the fourth sum (S3) is the greatest of the first to fourth sums;
and
(ii) the third sum (S2) is the greatest and the fourth sum (S3) is the second greatest of the first to fourth sums.

6 more elements

A people scene detecting method for detecting a people scene in an input image, comprising the steps of:

dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
detecting the following values for each block,
(a) an average value AV which is an average value of the colors in each block, and expressed by
AV.sub.13 YYx!y!, AV.sub.-- RYx!y!, AV.sub.-- BYx!y!
wherein AV.sub.13 YYx!y! is an average luminance, AV.sub.-- RYx!y! is an average color difference R-G, and AV.sub.-- BYx!y! is an average color difference B-G,
(b) a WP value which is the color of a pixel where the luminance is the highest in each block, and expressed by WP.sub.-- YYx!y!, WP.sub.-- RYx!y!, WP.sub.-- BYx!y!
wherein WP.sub.-- YYx!y! is a highest luminance, WP.sub.-- RYx!y! is a highest color difference R-G, and WP.sub.-- BYx!y! is a highest color difference B-G,
(c) a BP value which is the color of a pixel where the luminance is the lowest in each block, and expressed by
BP.sub.-- YYx!y!, BP.sub.-- RYx!y!, BP.sub.-- BYx!y!
wherein BP.sub.-- YYx!y! is a lowest luminance, BP.sub.-- RYx!y! is a lowest color difference R-G, and BP.sub.-- BYx!y! is a lowest color difference B-G;
selecting skin color blocks which satisfy a condition that the WP value and the average value AV are both in a second quadrant of a simple chromaticity diagram;
calculating averages av.sub.-- ry and av.sub.-- by of RY and BY, respectively, in accordance with the following equations, ##EQU20## wherein hada.sub.-- num is the number of selected skin color blocks;
calculating variances Srr and Sbb of RY and BY, respectively, in accordance with the following equations, ##EQU21## calculating covariance Srb in accordance with the following equation, ##EQU22## calculating a matrix S in accordance with the following equation, ##EQU23## calculating a proper vector, tan 2.theta., in accordance with the following equation, ##EQU24## where .theta. is an angle formed by a hue line that is the principle component axis and the positive BY axis;
detecting that the input image is a people scene when angle .theta. is in a range from .pi./4 to .pi. in the simple color difference chromaticity diagram.

11 more elements

A people scene detector for detecting a people scene in an input image, comprising:

dividing means for dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
detecting means for detecting the following values for each block,
(a) an average value AV which is an average value of the colors in each block, and expressed by
AV.sub.13 YYx!y!, AV.sub.-- RYx!y!, AV.sub.-- BYx!y!
wherein AV.sub.-- YYx!y! is an average luminance, AV.sub.-- RYx!y! is an average color difference R-G, and AV.sub.-- BYx!y! is an average color difference B-G,
(b) a WP value which is the color of a pixel where the luminance is the highest in each block, and expressed by
WP.sub.-- YYx!y!, WP.sub.-- RYx!y!, WP.sub.-- BYx!y!
wherein WP.sub.-- YYx!y! is a highest luminance, WP.sub.-- RYx!y! is a highest color difference R-G, and WP.sub.-- BYx!y! is a highest color difference B-G,
(c) a BP value which is the color of a pixel where the luminance is the lowest in each block, and expressed by
BP.sub.-- YYx!y!, BP.sub.-- RYx!y!, BP.sub.-- BYx!y!
wherein BP.sub.-- YYx!y! is a lowest luminance, BP.sub.-- RYx!y! is a lowest color difference R-G, and BP.sub.-- BYx!y! is a lowest color difference B-G;
block selecting means for selecting skin color blocks which satisfy that the WP value and the average value AV are both in a second quadrant of a simple chromaticity diagram;
average calculation means for calculating averages av.sub.-- ry and av.sub.-- by of RY and BY, respectively, in accordance with the following equations, ##EQU25## wherein hada.sub.-- num is the number of selected skin color blocks;
variance calculation means for calculating variances Srr and Sbb of RY and BY, respectively, in accordance with the following equations, ##EQU26## covariance calculation means for calculating covariance Srb in accordance with the following equation, ##EQU27## matrix calculation means for calculating a matrix S in accordance with the following equation, ##EQU28## proper vector calculation means for calculating a proper vector, tan 2.theta., in accordance with the following equation,
tan 2.theta.=2Srb/(Sbb-Srr)
where .theta. is an angle formed by a hue line that is the principle component axis and the positive BY axis;
people scene detecting means for detecting that the input image is a people scene when angle .theta. is in a range from .pi./4 to .pi. in the simple color difference chromaticity diagram.

14 more elements

A gradation correction apparatus for correcting the gradation of luminance in an input image comprising:

a scene information determination means for determining a scene information of the input image, said scene information determination means comprising a backlit scene detector for detecting a backlit scene in the input image, said backlit scene detector comprising:
dividing means for dividing the input image in to a plurality of blocks aligned in a plurality of columns and a plurality of rows;
luminance detecting means for detecting a luminance of each block;
dark block detecting means for detecting dark blocks among the plurality of blocks, each dark block having luminance which is smaller than a predetermined luminance;
dark block processing means for counting a first number of continuous dark blocks in each row which extend from a right side edge of the input image and obtaining a first sum (S0) of the first number of continuous dark blocks from all the rows, for counting a second number of continuous dark blocks in each row which extend from a left side edge of the input image and obtaining a second sum (S1) of the second number of continuous small blocks from all the rows, for counting a third number of continuous dark blocks in each column which extend from a top side edge of the input image and obtaining a third sum (S2) of the third number of continuous small blocks from all the columns, and for counting a third number of continuous dark blocks in each column which extend from a bottom side edge of the input image and obtaining a fourth sum (S3) of the fourth number of continuous small blocks from all the rows;
and
backlit scene detecting means for detecting that the input image is a backlit scene when either one of the following two conditions is satisfied:
(i) the fourth sum (S3) is the greatest of the first to fourth sums;
and
(ii) the third sum (S2) is the greatest and the fourth sum (S3) is the second greatest of the first to fourth sums;
a degree of backlit scene brightness information determination means for determining a degree of backlit scene brightness in the input image detected as a backlit scene;
a backlit basic density gradation table generating means for generating a backlit basic density gradation table;
a table modification means for modifying the backlit basic density gradation table based on said degree of backlit scene brightness to generate a backlit modified density gradation table;
a luminance table conversion means for converting said backlit modified density gradation table to a luminance table;
and
a gradation correction means for correcting the luminance gradation of the input image using said luminance table.

13 more elements

A gradation correction apparatus for correcting the gradation of luminance in an input image, comprising:

a scene information determination means for determining a scene information of the input image, said scene information determination means comprising a people scene detector for detecting a people scene in the input image, said people scene detector comprising:
dividing means for dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
detecting means for detecting the following values for each block,
(a) an average value AV which is an average value of the colors in each block, and expressed by
AV.sub.13 YYx!y!, AV.sub.-- RYx!y!, AV.sub.-- BYx!y!
wherein AV.sub.13 YYx!y! is an average luminance, AV.sub.-- RYx!y! is an average color difference R-G, and AV.sub.-- BYx!y! is an average color difference B-G,
(b) a WP value which is the color of a pixel where the luminance is the highest in each block, and expressed by
WP.sub.-- YYx!y!, WP.sub.-- RYx!y!, WP.sub.-- BYx!y!
wherein WP.sub.-- YYx!y! is a highest luminance, WP.sub.-- RYx!y! is a highest color difference R-G, and WP.sub.-- BYx!y! is a highest color difference B-G,
(c) a BP value which is the color of a pixel where the luminance is the lowest in each block, and expressed by
BP.sub.-- YYx!y!, BP.sub.-- RYx!y!, BP.sub.-- BYx!y!
wherein BP.sub.-- YYx!y! is a lowest luminance, BP.sub.-- RYx!y! is a lowest color difference R-G, and BP.sub.-- BYx!y! is a lowest color difference B-G;
block selecting means for selecting skin color blocks which satisfy a condition in which the WP value and the average value AV are both in a second quadrant of a simple chromaticity diagram;
average calculation means for calculating averages av.sub.-- ry and av.sub.-- by of RY and BY, respectively, in accordance with the following equations, ##EQU34## wherein hada.sub.-- num is the number of selected skin color blocks;
variance calculation means for calculating variances Srr and Sbb of RY and BY, respectively, in accordance with the following equations, ##EQU35## covariance calculation means for calculating covariance Srb in accordance with the following equation, ##EQU36## matrix calculation means for calculating a matrix S in accordance with the following equation, ##EQU37## proper vector calculation means for calculating a proper vector, tan 2.theta., in accordance with the following equation, ##EQU38## where .theta. is an angle formed by a hue line that is the principle component axis and the positive BY axis;
people scene detecting means for detecting that the input image is a people scene when angle .theta. is in a range from .pi./4 to .pi. in the simple color difference chromaticity diagram,
a degree of people scene brightness information determination means for determining a degree of people scene brightness in the input image detected as a people scene;
a people basic density gradation table generating means for generating a people basic density gradation table;
a table modification means for modifying the people basic density gradation table based on said degree of people scene brightness to generate a people modified density gradation table;
a luminance table conversion means for converting said people modified density gradation table to a luminance table;
and
a gradation correction means for correcting the luminance gradation of the input image using said luminance table.

19 more elements

A gradation correction apparatus for correcting the gradation of luminance in an input image comprising:

a scene information determination means for determining a scene information of the input image, said scene information determination means comprising a backlit scene detector for detecting a backlit scene in the input image, said backlit scene detector comprising:
dividing means for dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
luminance detecting means for detecting a luminance of each block;
dark block detecting means for detecting dark blocks among the plurality of blocks, each dark block having a luminance which is smaller than a predetermined luminance;
dark block processing means for counting a first number of continuous dark blocks in each row which extend from a right side edge of the input image and obtaining a first sum (S0) of the first number of continuous dark blocks from all the rows, for counting a second number of continuous dark blocks in each row which extend from a left side edge of the input image and obtaining a second sum (S1) of the second number of continuous dark blocks from all the rows, for counting a third number of continuous dark blocks in each column which extend from a top side edge of the input image and obtaining a third sum (S2) of the third number of continuous dark blocks from all the columns, and for counting a fourth number of continuous dark blocks in each column which extend from a bottom side edge of the input image and obtaining a fourth sum (S3) of the fourth number of continuous dark blocks from all the rows;
and
backlit scene detecting means for detecting that the input image is a backlit scene when either one of the following two conditions is satisfied:
(i) the fourth sum (S3) is the greatest of the first to fourth sums;
and
(ii) the third sum (S2) is the greatest and the fourth sum (S3) is the second greatest of the first to fourth sums;
a degree of backlit scene brightness information determination means for determining a degree of backlit scene brightness in the input image detected as a backlit scene;
a backlit basic density gradation table generating means for generating a backlit basic density gradation table;
a backlit luminance table conversion means for converting said backlit basic density gradation table to a backlit luminance table;
and
a table modification means for modifying the backlit luminance table based oh said degree of backlit scene brightness to generate a backlit modified luminance table;
a gradation correction means for correcting the luminance gradation of the input image using said modified luminance table.

13 more elements

A gradation correction apparatus for correcting the gradation of luminance in an input image, comprising:

a scene information determination means for determining a scene information of the input image, said scene information determination means comprising a people scene detector for detecting a people scene in the input image, said people scene detector comprising:
dividing means for dividing the input image into a plurality of blocks aligned in a plurality of columns and a plurality of rows;
detecting means for detecting the following values for each block,
(a) an average value AV which is an average value of the colors in each block, and expressed by
AV.sub.13 YYx!y!, AV.sub.-- RYx!y!, AV.sub.-- BYx!y!
wherein AV.sub.13 YYx!y! is an average luminance, AV.sub.-- RYx!y! is an average color difference R-G, and AV.sub.-- BYx!y! is an average color difference B-G,
(b) a WP value which is the color of a pixel where the luminance is the highest in each block, and expressed by
WP.sub.-- YYx!y!, WP.sub.-- RYx!y!, WP.sub.-- BYx!y!
wherein WP.sub.-- YYx!y! is a highest luminance, WP.sub.-- RYx!y! is a highest color difference R-G, and WP.sub.-- BYx!y! is a highest color difference B-G,
(c) a BP value which is the color of a pixel where the luminance is the lowest in each block, and expressed by
BP.sub.-- YYx!y!, BP.sub.-- RYx!y!, BP.sub.-- BYx!y!
wherein BP.sub.-- YYx!y! is a lowest luminance, BP.sub.-- RYx!y! is a lowest color difference R-G, and BP.sub.-- BYx!y! is a lowest color difference B-G;
block selecting means for selecting skin color blocks which satisfy a condition in which the WP value and the average value AV are both in a second quadrant of a simple chromaticity diagram,
average calculation means for calculating averages av.sub.-- ry and av.sub.-- by of RY and BY, respectively, in accordance with the following equations, ##EQU43## wherein hada.sub.-- num is the number of selected skin color blocks;
variance calculation means for calculating variances Srr and Sbb of RY and BY, respectively, in accordance with the following equations, ##EQU44## covariance calculation means for calculating covariance Srb in accordance with the following equation, ##EQU45## matrix calculation means for calculating a matrix S in accordance with the following equation, ##EQU46## proper vector calculation means for calculating a proper vector, tan 2.theta., in accordance with the following equation, ##EQU47## where .theta. is an angle formed by a hue line that is the principle component axis and the positive BY axis;
people scene detecting means for detecting that the input image is a people scene when angle .theta. is in a range from .pi./4 to .pi. in the simple color difference chromaticity diagram,
a degree of people scene brightness information determination means for determining a degree of people scene brightness in the input image detected as a people scene;
a people basic density gradation table generating means for generating a people basic density gradation table;
a luminance table conversion means for converting said people basic density gradation table to a people luminance table;
and
a table modification means for modifying the people luminance table based on said degree of people scene brightness to generate a people modified luminance table;
a gradation correction means for correcting the luminance gradation of the input image using said people modified luminance table.

19 more elements

Backlit scene and people scene detecting method and apparatus and a gradation correction apparatus

Life of this patent

Backlit scene and people scene detecting method and apparatus and a gradation correction apparatus

Life of this patent

Abstract

Claims23