Digital image processing method having an exposure correction based on recognition of areas corresponding to the skin of the photographed subject

• US 7,778,483 B2
• Filed: 12/03/2007
• Issued: 08/17/2010
• Est. Priority Date: 05/19/2003
• Status: Active Grant

• Current Assignee: Stmicroelectronics France (STMicroelectronics NV)
• Original Assignee: STMicroelectronics SA (STMicroelectronics NV), STMicroelectronics SRL (STMicroelectronics NV)
• Inventors: Plaza, Laurent, Castorina, Alfio, Messina, Giuseppe, Battiato, Sebastiano
• Primary Examiner(s): PERUNGAVOOR, SATHYANARAYA V
• Application Number:
  US11/949,709
• Publication Number:
  US 20080089583A1
• Time in Patent Office: 988 Days
• Field of Search: 382/274
• US Class Current: 382/274
• CPC Class Codes: G06T 5/40 using histogram techniques
• CPC Class Codes: G06T 5/90 Dynamic range modification of images or parts thereof
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching
• CPC Class Codes: H04N 1/6027 Correction or control of colour gradation or colour contrast H04N1/6058 takes precedence

Abstract

A digital image processing method includes extracting chromatic information of an image taken by an image taking device and related to a human subject; detecting visually interesting regions; and exposure correcting of the taken image by normalizing a grey scale of the taken image based on the visually interesting regions. Advantageously, the method includes recognizing areas corresponding to the skin of the subject, these areas being used as the visually interesting regions for the exposure correction step.

Claims

1. A computer implemented digital image processing method, comprising:

extracting chromatic information of a Bayer type image taken by an image taking device and related to a human subject;

sub-sampling the Bayer type image according to G*=(G1+G2)/2, wherein G1 and G2 are first and second green channels of the Bayer type image to produce an RGB image;

constructing a chrominance slicing histogram of the RGB image using normalized channels r and g of the type;

r=R/(R+G*+B)

g=G*/(R+G*+B),

R, G and B being red, green and blue values of each pixel of the RGB image;

detecting visually interesting regions in the RGB image by recognizing areas corresponding to skin of the subject using the chrominance slicing histogram, wherein the recognized areas are the visually interesting regions; and

correcting an exposure of the Bayer type image or the RGB image by normalizing a grey scale of the Bayer type image or the RGB image based on said visually interesting regions.

2. A digital image processing method according to claim 1, wherein said exposure correction of said Bayer type image uses:

a simulated response function of a type;

$f<br><br>\left(q\right)=\frac{255}{{\left(1+{\mathrm{e<br><br>}}^{-\left(\mathrm{Aq}\right)}\right)}^C}$

A and C being predetermined control parameters and q being a light quantity value expressed in base 2 logarithmic units; and

a grey average level (avg) calculated on said visually interesting regions, in order to calculate a distance Δ

of an ideal exposure situation using;

Δ

=f<sup>−

1</sup>(128)−

f<sup>−

1</sup>(avg)

and in order to change a luminance value Y(x, y) of a pixel with position (x, y) in;

Y′

(x, y)f(f<sup>−

1</sup>(Y(x, y))+Δ

).

3. A digital image processing method according to claim 1, further comprising a final color reconstruction step, wherein said final color reconstruction step provides that a grey value I(x, y) of a pixel with position (x, y) is changed in:

I′

(x, y)=f(f<sup>−

1</sup>(I(x, y))+Δ

),

where Δ

is a distance of an ideal exposure situation.

4. A computer implemented digital image processing method, comprising:

extracting chromatic information from a Bayer type image of a human subject;

sub-sampling the Bayer type image according to G*=(G1+G2)/2, wherein G1 and G2 are first and second green channels of the Bayer type image to produce an RGB image;

constructing a chrominance slicing histogram of the RGB image using normalized channels r and g of the type;

r=R/(R+G*+B)

g=G*/(R+G*+B),

R, G and B being red, green and blue values of each pixel of the RGB image;

detecting, based on the chrominance slicing histogram, which areas of the RGB image correspond to skin of the subject; and

normalizing grey scale values of the Bayer type image or the RGB image based on the areas that are detected as corresponding to the skin of the subject.

5. The method of claim 4, wherein the normalizing step performs exposure correction of the Bayer type image that includes:

using a simulated response function of a type;

$f<br><br>\left(q\right)=\frac{255}{{\left(1+{\mathrm{e<br><br>}}^{-\left(\mathrm{Aq}\right)}\right)}^C}$

A and C being predetermined control parameters and q being a light quantity value expressed in base 2 logarithmic units; and

calculating a grey average level (avg) of the areas corresponding to the skin;

calculating a distance Δ

of an ideal exposure situation using;

Δ

f<sup>−

1</sup>(128)−

f<sup>−

1</sup>(avg); and

changing a luminance value Y(x, y) of a pixel with position (x, y) in;

Y′

(x, y)=f(f<sup>−

1</sup>(Y(x, y))+Δ

).

6. The method of claim 4, further comprising a final color reconstruction step that changes a grey value I(x, y) of a pixel with position (x, y) using:

I′

(x, y)=f(f<sup>−

1</sup>(I(x, y))+Δ

),

where Δ

is a distance of an ideal exposure situation.

7. A digital image processor, comprising:

means for extracting chromatic information from a Bayer type image of a human subject;

means for sub-sampling the Bayer type image according to G*=(G1+G2)/2, wherein G1 and G2 are first and second green channels of the Bayer type image to produce an RGB image;

means for constructing a chrominance slicing histogram of the RGB image using normalized channels r and g of the type;

r=R/(+G*+B)

g=G*/(R+G*+B)

R, G and B being red, green and blue values of each pixel of the RGB images;

means for detecting, based on the chrominance slicing histogram, which areas of the RGB image correspond to skin of the subject; and

means for normalizing grey scale values of the Bayer type image or the RGB image based on the areas that are detected as corresponding to the skin of the subject.