Image processing method, facial region extraction method, and apparatus therefor

• US 6,816,611 B1
• Filed: 05/26/1999
• Issued: 11/09/2004
• Est. Priority Date: 05/29/1998
• Status: Expired due to Fees

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Hagiwara, Masafumi, Kimura, Toshihiro, Tadokoro, Yoshihisa, Yokoo, Yuuki, Kato, Masami, Sakakibara, Ken
• Primary Examiner(s): Ahmed, Samir
• Assistant Examiner(s): BHATNAGAR, ANAND P
• Application Number:
  US09/318,818
• Time in Patent Office: 1,994 Days
• Field of Search: 382/117, 382/118, 382/164, 382/165, 382/167, 382/168, 382/171, 348/77, 348/143, 348/156, 340/5.53, 340/5.83, 340/672, 713/186, 707/6, 902/3, 902/6, 358/520, 358/521, 358/522, 358/453, 702/180
• US Class Current: 382/165
• CPC Class Codes: G06V 10/507 Summing image-intensity values; Histogram projection analysis
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching

Abstract

A flesh tone region detection processor (12) detects a flesh tone region from a color still image, and a projective distribution generator (13) generates the projective distribution of the detected flesh tone region. A facial region search processor (14) searches the projective distribution for a parabola using, e.g., a genetic algorithm, and extracts a facial region candidate from the position of the found parabola. A knowledge processor (15) checks based on the aspect ratio of the region, matching with facial patterns prepared in advance, or the presence/absence of a region corresponding to hair if the extracted facial region candidate is a facial region. The user can set the processing time, the number of persons to be detected, and the face area ratio as a search end condition.

Claims

1. A facial region extraction method for extracting a facial region of a person from a color image, comprising:

a detection step of detecting a flesh tone region;

a generation step of generating a projective distribution of the detected flesh tone region;

an obtaining step of obtaining a graph of the projective distribution;

a search step of searching a parabola corresponding to the graph;

an extraction step of extracting a facial region candidate from a position of the parabola found by the search; and

a determination step of determining if the extracted facial region candidate is a facial region,

wherein said search step includes a step of searching for a parabola by a genetic algorithm that matches the projective distribution against the parabola with changing parameters of the parabola, and

wherein, at said search step, a slope and a vertex defining a parabola are used as a chromosome in the genetic algorithm and the search of the parabola is performed with crossing over or mutating the chromosome.

2. The method according to claim 1, wherein the detection step uses hue and color difference components, and a region having predetermined hue and color difference component is detected as the flesh tone region.

3. The method according to claim 1, wherein the extraction step uses a parabola obtained from a projective distribution in only one axis direction of a coordinate system.

4. The method according to claim 1, wherein the extraction step uses two parabolas obtained from projective distributions in two axis directions of a coordinate system.

5. The method according to claim 1, wherein the determination step includes a step of determining if the facial region candidate is a facial region using an aspect ratio of the facial region candidate.

6. The method according to claim 1, wherein the determination step includes a step of determining if the facial region candidate is a facial region using a neural network that has learned facial patterns in advance.

7. The method according to claim 6, wherein the neural network has learned face data in a plurality of directions as teacher images.

8. The method according to claim 1, wherein the determination step includes a step of determining if the facial region candidate is a facial region based on the presence/absence of a hair region in the facial region candidate.

9. A facial region extraction apparatus for extracting a facial region of a person from a color image, comprising:

detection means for detecting a flesh tone region;

generation means for generating a projective distribution of the detected flesh tone region;

obtaining means for obtaining a graph of the projective distribution;

search means for searching a parabola corresponding to the graph;

extraction means for extracting a facial region candidate from a position of the parabola found by the search; and

determination means for determining if the extracted facial region candidate is a facial region,

wherein said search means searches for a parabola by a genetic algorithm that matches the projective distribution against the parabola with changing parameters of the parabola, and

wherein said search means uses a slope and a vertex defining a parabola as a chromosome in the genetic algorithm and searches the parabola with crossing over or mutating the chromosome.

10. The apparatus according to claim 9, wherein said detection means uses hue and color difference components, and a region having predetermined hue and color difference components is detected as the flesh tone region.

11. The apparatus according to claim 9, wherein said extraction means uses a parabola obtained from a projective distribution in only one axis direction of a coordinate system.

12. The apparatus according to claim 9, wherein said extraction means uses two parabolas obtained from projective distributions in two axis directions of a coordinate system.

14. The apparatus according to claim 9, wherein said determination means determines if the facial region candidate is a facial region using a neural network that has learned facial patterns in advance.

15. The apparatus according to claim 14, wherein the neural network has learned face data in a plurality of directions as teacher images.

16. The apparatus according to claim 9, wherein said determination means determines if the facial region candidate is a facial region based on the presence/absence of a hair region in the facial region candidate.

13. The apparatus according to claim wherein said determination means determines if the facial region candidate is a facial region using an aspect ratio of the facial region candidate.

17. A storage medium storing a program that makes a computer function as:

means for detecting a flesh tone region;

generation means for generating a projective distribution of the detected flesh tone region;

obtaining means for obtaining a graph of the projective distribution;

search means for searching a parabola corresponding to the graph;

extraction means for extracting a facial region candidate from a position of the parabola found by the search; and

determination means for determining if the extracted facial region candidate is a facial region,

wherein said search means searches for a parabola by a genetic algorithm that matches the projective distribution against the parabola with changing parameters of the parabola, and

wherein said search means uses a slope and a vertex defining a parabola as a chromosome in the genetic algorithm and searches the parabola with crossing over or mutating the chromosome.