JP2004355084A - Image processor and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To quickly extract a specific area from an input image. <P>SOLUTION: In image processing for extracting an area surrounded by a shape expressed with a mathematical formula described by a parameter from an image, an image is inputted, and one point in the input image is set as an initial point, and a plurality of straight lines passing the initial point are set. As for pixels on the plurality of straight lines, the boundary points of an area where pixels having specific characteristics are continuously present are detected based on pixel data, and a parameter describing the mathematical formula is calculated so that it can pass the boundary points, and a shape described by the parameter is outputted. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

で表される楕円曲線を抽出曲線として設定し、パラメータ算出部２１２が、ピクセル選択部２０６で求められた４個の境界点１８のピクセルの座標値を用いて、それらの点１８を通る楕円曲線の方程式を解き、式（１）のパラメータａ， ｂ， ｘ０， ｙ０の値を数値解析を用いて求める。こうして、画像中から式（１）の楕円曲線で記述できる領域を抽出する。適当なパラメータ群が求まった場合、パラメータ曲線出力部２１４は、そのパラメータ値で記述される楕円曲線２０を出力する。
【００１６】
抽出直線の数は、抽出直線設定部２０４においてあらかじめ設定しておく。たとえば、抽出直線の数を３と設定してあれば、抽出直線設定部２０４は３本の抽出直線を設定し、ピクセル特徴算出部２０６は、これらの３本の抽出直線について特徴量を算出し、ピクセル選択部２０８は、３本の抽出直線について肌色と非肌色との境界点１８を求める。パラメータ算出部２１２は、その結果を基に楕円曲線を記述するパラメータ値を計算する。
【００１７】
図６は、ＣＰＵ１００による領域抽出のフローチャートである。まず、顔画像を入力する（Ｓ１００）。次に、入力された顔画像に対して、操作者により設定された初期点１４を入力し（Ｓ１０２）、操作者により設定された初期点１４を通る複数（たとえば２本）の直線１６を入力する（Ｓ１０４）。次に、これらの直線上のピクセルの特徴量を算出し（Ｓ１０６）、算出した特徴量を基に肌色の特徴を持つピクセルを選択する。そして、直線１６にそって非肌色と肌色との境界点１８を求める（Ｓ１０８）。次に、境界点１８を通る適当な形状（たとえば楕円曲線）を表わす数式（たとえば式（１））のパラメータ値を数値解析により計算する（Ｓ１１０）。こうして、画像中から式（１）の楕円曲線で記述できる領域を抽出する。そして、それらのパラメータ値で記述される楕円曲線２０を出力する（Ｓ１１２）。
【００１８】
実施形態２．
図７は、実施形態２の領域抽出処理を説明するための図である。ここでは、実施形態１の手法で求めた楕円曲線を記述するパラメータ群（第１曲線パラメータ群）の妥当性を評価する。このため、求まった曲線パラメータで記述される領域の内部において初期点１４とは異なる第２点（評価点という）２４を設定し、初期点１４の場合と同様に、第２点２４を通る２以上の直線２６を設定して、適当な曲線記述パラメータ群（第２曲線パラメータ群）を算出する。そして、両者を比較して曲線パラメータ群の妥当性を評価する。１つの点（初期点１４）からパラメータ値を求めるだけではなく、第２の点（評価点）２４を用意して、初期点１４を用いて求めたパラメータ値の評価を行うので、精度の高いパラメータ計算が可能となる。
【００１９】
図８は、図１に示した画像処理装置における領域抽出処理の機能ブロック図である。画像中からあるパラメータ曲線で記述できる領域を抽出する際に、まず、実施形態１と同様に、画像中の任意の点（初期点１４）を通る２本以上の線１６を設定し、線上の画素（ピクセル）の特徴量を元に、肌色と非肌色の境界のピクセル（境界点１８）を選択し、選択したピクセル群を基に適当な楕円パラメータを計算する。
【００２０】
次に、評価点設定部２１６において、操作者は、パラメータ算出部２１２において求まった楕円曲線２０の内部の、初期点１４以外の１点を評価点２４（図７参照）として設定し、抽出直線設定部２０４において、操作者は、評価点２４を通る２本の直線２６を設定する。以下、初期点１４を通る直線の場合と同様に、ピクセル特徴算出部２０６とピクセル選択部２０８は、評価点を通る２直線の上の画素（ピクセル）の特徴量を元に、特定の特徴を持つピクセル（境界点２８）を選択する。次に、パラメータ算出部２１２は、それらの４個の境界点２８をとおる楕円曲線を記述するパラメータ群を求める。なお、評価点２４を通る抽出曲線の数は、初期点を通る抽出直線の場合と同様に、３本以上に増やしてもよい。
【００２１】
パラメータ評価部２２０は、初期点１４を用いて求まったパラメータ群の値と評価点２４を用いて求まったパラメータ群の値との最小自乗誤差を求め、パラメータ閾値設定部２１８において設定されているパラメータ閾値と比較する。パラメータ曲線出力部２１４は、最小自乗誤差がパラメータ閾値以下であると評価された場合に、初期点１４を用いて求まったパラメータ群で記述される楕円曲線２０を出力する。
【００２２】
図９は、実施形態２における画像処理のフローチャートである。まず、顔画像を入力する（Ｓ１００）。次に、入力された顔画像に対して、操作者により設定された初期点１４を入力し（Ｓ１０２）、操作者により設定された初期点１４を通る複数（たとえば２本）の直線１６を入力する（Ｓ１０４）。次に、これらの直線上のピクセルの特徴量を算出し（Ｓ１０６）、算出した特徴量を基に肌色の特徴を持つピクセルを選択する。そして、非肌色と肌色との境界点１８を求める（Ｓ１０８）。次に、境界点１８を通る適当な形状（たとえば楕円曲線）を表わす数式（たとえば式（１））のパラメータ値を数値解析により計算する（Ｓ１１０）。こうして、画像中から式（１）の楕円曲線で記述できる領域を抽出する。以上の処理は、実施の形態１と同様である。
【００２３】
次に、求まった楕円曲線２０の内部の、操作者により設定された初期点１４以外の１点を評価点２４として入力し（Ｓ１１２）、操作者により設定された、評価点２４を通る複数の直線２６を設定する（Ｓ１１４）。以下、初期点１４を通る直線の場合と同様に、これらの直線２６上のピクセルの特徴量を算出し（Ｓ１１６）、算出した特徴量を基に肌色の特徴を持つピクセルを選択して、直線２６にそって非肌色と肌色との境界点２８を求める（Ｓ１１８）。次に、それらの４個の抽出点２８を基に、曲線パラメータの値を求める（Ｓ１２０）。そして、曲線パラメータの評価のため、初期点１４を用いて求めたパラメータ値と評価点２４を用いて求めたパラメータ値との最小自乗誤差を求め、パラメータしきい値と比較する（Ｓ１２２）。最小自乗誤差がパラメータしきい値以下であると評価した場合に、初期点１４において求めたパラメータ値で記述される楕円曲線を出力する（Ｓ１２４）。
【００２４】
実施形態３．
画像には複数の人の顔が含まれていることがある。そのような場合、以下に説明するように、画像を複数の領域に分割して、それぞれ楕円領域を抽出する。これにより、２人の顔の検出を、少ない計算処理で高速に行う。
【００２５】
図１０は、１枚の顔画像４０に２人の顔の肌色領域４２、４４が分離して存在する場合を示す。まず、実施の形態２の画像処理と同様に、操作者が、初期点４６を設定し、初期点４６を通る２本の抽出直線４８を設定する。初期点４４は、２つの肌色領域４２，４４の境界線の上に設定する。そして、これらの直線の上の画素（ピクセル）の特徴量を求め、特定の特徴を持つピクセル（袖出点５０）を選択する。そして、袖出点５０を基に、適当な楕円曲線を記述するパラメータを計算して、楕円曲線５２を求める。
【００２６】
次に、領域分割の評価のために、操作者は楕円曲線５２内に、初期点４６と異なる位置に評価点５４を設定し、評価点５４を通る２本の抽出直線５６を設定する。次に、これらの直線の上の画素（ピクセル）の特徴量を元に、特定の特徴を持つピクセル（袖出点５８）を選択する。そして、袖出点５８から、楕円曲線を記述するパラメータ値を計算する。楕円曲線６０は、こうして得られた楕円を示す。
【００２７】
次に、初期点４６を用いて求めたパラメータ値と評価点５４を用いて求めたパラメータ値との最小自乗誤差を求め、領域分割しきい値と比較してパラメータ値の評価を行う。領域分割しきい値を越えた場合、図１１に示すように、初期点４６の左右に分割時初期点６２を設定する。そして、各初期点６２について、分割時初期点６２を通る２本の直線６４を設定し、初期点４６の場合と同様に、図示しないが、直線にそって抽出点を求め、それらの抽出点を基に楕円領域を求める。
【００２８】
また、図１２は、１枚の顔画像７０に２人の顔の肌色領域７２、７４が重なって存在する場合を示す。まず、実施の形態２の画像処理と同様に、操作者が、初期点７６を設定し、初期点７６を通る複数本の抽出直線７８を設定する。初期点４４は、２つの肌色領域７２，７４の境界線の上に設定する。そして、これらの直線７８の上の画素（ピクセル）の特徴量を元に、特定の特徴を持つピクセル（袖出点８０）を選択する。そして、袖出点８０から、適当な楕円曲線を記述するパラメータ値を計算して、楕円曲線８２を求める。
【００２９】
次に、領域分割の評価のために、操作者は、楕円曲線８２内に、初期点７６と異なる位置に評価点８４を設定し、評価点８４を通る２本の抽出直線８６を設定する。次に、これらの直線８６の上の画素（ピクセル）の特徴量を元に、特定の特徴を持つピクセル（抽出点８８）を選択する。そして、抽出点８８から、適当な楕円曲線を記述するパラメータ値を計算する。楕円曲線９０は、こうして得られた楕円を示す。
【００３０】
次に、初期点７６を用いて求めたパラメータと評価点８４を用いて求めたパラメータとの最小自乗誤差を求め、領域分割しきい値と比較して楕円パラメータの評価を行う。領域分割しきい値を越えた場合、図１３に示すように、初期点７６の左右に分割時初期点９２を設定する。分割時初期点の位置は肌色領域７２の左側と肌色領域７４の右側とし、直線９４は、２つの肌色領域にわたらないように設定する。そして、各分割時初期点９２について、分割時初期点９２を通る２本の直線９４を設定し、初期点７６と同様に、図示しないが、直線に沿って抽出点を求め、それらの抽出点を基に楕円領域を求める。
【００３１】
図１４は、図１に示した画像処理装置における画像処理の機能ブロック図である。画像中からあるパラメータ曲線で記述できる領域を抽出する際に、まず、実施形態１と同様に、ブロック２００〜２１２において、画像中の任意の点（初期点）を通る２以上の線の上の画素（ピクセル）の特徴量を元に、特定の特徴を持つピクセルを選択し、肌色と非肌色との境界を求め、選択したピクセル群から適当なパラメータ値を計算する。
【００３２】
パラメータ評価部２２０が、領域分割しきい値設定部２２２により設定された領域分割しきい値を超えたと評価する場合（たとえば図１０や図１１の場合）、分割時初期点設定部２２４は、図１２と図１３の例に示すように、分割時初期点６２，９２を初期点４０，７０の左右にそれぞれ設定し、分割時初期点６２，９２ごとに、実施形態１と同様な計算を行って、楕円領域を抽出する。
【００３３】
図１５は、実施の形態３における画像処理のフローチャートである。まず、顔画像を入力する（Ｓ１００）。次に、入力された顔画像に対して、操作者により設定された初期点４６，７６を入力し（Ｓ１０２）、操作者により設定された初期点４６，７６を通る複数（たとえば２本）の直線４８，７８を入力する（Ｓ１０４）。次に、これらの直線上のピクセルの特徴量を算出し（Ｓ１０６）、算出した特徴量を基に肌色の特徴を持つピクセルを選択する。そして、非肌色と肌色との境界点５０、８０を求める（Ｓ１０８）。次に、境界点５０、８０を通る適当な曲線（楕円）を表わす数式（たとえば式（１））のパラメータ値を数値解析により計算する（Ｓ１１０）。こうして、画像中から式（１）の楕円曲線で記述できる領域を抽出する。
【００３４】
次に、求まった楕円曲線の内部の、操作者により設定された初期点４６，７６以外の１点を評価点５４，８４として入力し（Ｓ１１２）、操作者により設定された、評価点５４，８４を通る複数の直線５６，８６を設定する（Ｓ１１４）。以下、初期点４６，７６を通る直線４８，７８の場合と同様に、これらの直線上のピクセルの特徴量を算出し（Ｓ１１６）、算出した特徴量を基に肌色の特徴を持つピクセルを選択する。そして、非肌色と肌色との境界点５０，８０を求める（Ｓ１１８）。次に、それらの４個の境界点５０，８０を基に、パラメータ値を求める（Ｓ１２０）。そして、パラメータ値の評価のため、初期点４６，７６を用いて求めたパラメータ値と評価点５４，８４を用いて求めたパラメータ値との最小自乗誤差を求め、領域分割用しきい値と比較する（Ｓ１２６）。
【００３５】
最小自乗誤差が領域分割用閾値を越えると評価された場合、初期点４６，７６の左右に分割時初期点６２，９２を設定し（Ｓ１２８）、各分割時初期点６２，９２を通る複数の直線６４，９４を設定する（Ｓ１３０）。次に、これらの直線６４，９４上のピクセルの特徴量を算出し（Ｓ１３２）、算出した特徴量を基に肌色の特徴を持つピクセルを選択する。そして、直線６４，９４にそって非肌色と肌色との境界点を求める（Ｓ１３４）。次に、境界点を通る適当な曲線（楕円）を表わす数式（たとえば式（１））のパラメータ値を数値解析により計算する（Ｓ１３６）。こうして、画像中から式（１）の楕円曲線で記述できる複数の領域を抽出する。
【００３６】
なお、以上の実施形態では、初期点、評価点、分割時初期点を同じ一枚の画像から選択しているが、例えば、入力画像が動画像の場合は、異なるフレームから選ぶことも可能である。
【００３７】
【発明の効果】
画像からの領域の抽出および領域の形状を記述する数式のパラメータ値の抽出が、少ない計算処理で可能である。
画像の中の、重なった複数の領域の個別抽出および個別パラメータ値抽出が、少ない計算処理で可能である。
【図面の簡単な説明】
【図１】実施の形態１における画像処理を説明する図
【図２】図１により説明した画像処理により得られた楕円領域を示す図
【図３】実施の形態１における画像処理の変形例を説明する図
【図４】画像処理装置のブロック図
【図５】実施の形態１の情報処理の機能ブロック図
【図６】実施の形態１の情報処理のフローチャート
【図７】実施の形態２における画像処理を説明する図
【図８】実施の形態２の情報処理の機能ブロック図
【図９】実施の形態２の情報処理のフローチャート
【図１０】実施の形態３における第１例の画像処理を説明する図
【図１１】実施の形態３における第１例の画像処理を説明する別の図
【図１２】実施の形態３における第２例の領域分割処理を説明する図
【図１３】実施の形態３における第２例の領域分割処理を説明する別の図
【図１４】実施の形態３の情報処理の機能ブロック図
【図１５】実施の形態３の情報処理のフローチャート
【符号の説明】
１０ 画像、 １２ 肌色領域、 １４ 初期点、 １６ 直線、 １８ 境界点、 ２０ 楕円曲線、 ２４ 評価点、 ２６ 直線、 ４６ 初期点、 ４８ 直線、 ５０ 境界点、 ５４ 評価点、 ５６ 直線、 ５８ 境界点、 ６２ 分割時初期点、 ６４ 直線、 ７６ 初期点、 ７８ 直線、 ８０ 境界点、 ８４ 評価点、 ８６ 直線、 ８８ 境界点、 ９２ 分割時初期点、 ９４ 直線、 １００ ＣＰＵ、 １０２ ＲＯＭ、 １０６ フレキシブルディスク、 １０８ ディスプレイ装置、 １１０ ハードデイスク、 １１４ ＣＤ−ＲＯＭ、 １２０ キーボード、 １２２ マウス。[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to image processing, and more particularly, to region extraction from an image.
[0002]
[Prior art]
As an image processing, a region having a specific property is extracted from an input image. For example, various techniques have been proposed for extracting a human face (skin color) region from a face image. For example, in the face extraction method described in Japanese Patent Application Laid-Open No. 8-63597, a skin color region is extracted from the entire color image, edge processing is performed, and a plurality of templates of various shapes of the face are used using the edge information. Is performed to extract a face area.
[0003]
In the face area extraction method described in JP-A-8-122944, after binarizing an image, area division is performed, a specific area is selected based on the feature amount of the area, and a specific area is selected from the selected area. Then, a shape pattern representing the contour of the face is extracted.
[0004]
Further, in the face region extraction device described in Japanese Patent Application Laid-Open No. H11-18502, after extracting a skin color region having a maximum area from the entire image which is a rectangular region, iterative processing of various elliptic parameters approximating the entire region is repeated. Then, an appropriate ellipse that approximates the region is obtained.
[0005]
[Patent Document 1]
JP-A-8-63597 [Patent Document 2]
Japanese Patent Application Laid-Open No. 8-122555 [Patent Document 3]
JP-A-11-18502
[Problems to be solved by the invention]
However, the conventional methods for extracting a human face (skin color) region from a face image have a problem that a computer load is high. For example, in the face extraction method described in Japanese Patent Application Laid-Open No. 8-63597, the amount of calculation is large because the entire image is processed, and the computer load is high.
[0007]
An object of the present invention is to rapidly extract a region having a specific property from an input image.
[0008]
[Means for Solving the Problems]
In the image processing according to the present invention, an image is input, one point in the input image is set as an initial point, and a plurality of straight lines passing through the initial point are set. Next, for the pixels on the plurality of straight lines, based on the pixel data, a boundary point of a region where pixels having specific characteristics are continuously present is detected, and the pixel passes through the boundary point. A parameter describing the above mathematical formula is calculated, and a shape described by the parameter is output. As a result, an area surrounded by the shape represented by the mathematical expression described by the parameter is extracted from the image.
[0009]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference symbols indicate the same or equivalent ones.
[0010]
Embodiment 1 FIG.
In the image processing apparatus according to the first embodiment, a region having a certain characteristic (for example, a skin color region) is extracted from a color image as a region represented by a curve described by a mathematical expression including a parameter. Here, in order to easily perform the extraction process, a plurality of (for example, two) lines passing through an arbitrary point (initial point) in the image are set, and the line is changed from the flesh color region to the non-skin color region along these straight lines. And a boundary point that changes from the non-skin color region to the skin color region. Here, from the viewpoint of extracting a region including the initial point when obtaining the boundary point, for example, it is effective to obtain the boundary point in a direction spreading in four directions from the initial point. Then, values of parameters (also referred to as curve parameters) describing a shape (for example, an elliptic curve) are obtained so as to pass through the obtained boundary points. The shape described by the obtained parameter values is an initial curve approximating the region having the above-mentioned certain characteristic. As described above, in the image area extraction, the calculation amount is largely reduced by calculating not on all pixels of the rectangular area but on pixels on two straight lines including the initial point.
[0011]
This region extraction processing will be described with reference to FIGS. As shown in FIG. 1, the image 10 includes a skin color area 12. An arbitrary point in the image 10 is set as an initial point 14, and two straight lines 16 passing through the initial point 14 are set. In the example of FIG. 1, the initial point 14 is at the center of the image. For each pixel on the straight line 16, a characteristic amount indicating a flesh color is obtained, and a boundary point 18 between the flesh color region and the non-skin color region is extracted based on the feature amount. (The boundary points are also referred to as extraction points 18.) In the example of FIG. 1, four extraction points 18 were obtained. A shape approximating the shape of the skin color region is defined as an elliptic curve. As shown in FIG. 2, a parameter value describing an elliptic curve 20 passing through these extraction points 18 is obtained, and a region represented by the elliptic curve 20 is extracted to approximate the skin color region 12.
[0012]
In the above example, the number of lines passing through the initial point is two, but the number of extracted straight lines may be three or more. For example, in the example shown in FIG. 3, an additional extraction line 16 'is set in addition to the two extraction lines 16 as shown in FIG. That is, the number of extracted straight lines is three. Therefore, the values of the parameters describing the elliptic curve are determined based on the six extraction points 18, 18 '. By performing parameter evaluation of the elliptic curve by increasing the number of extracted straight lines, more accurate parameter calculation can be performed. In this example, the position of the initial point 14 is shifted from the center of the screen 10.
[0013]
FIG. 4 illustrates an image processing apparatus according to the first embodiment. The configuration of this image processing apparatus is the same as that of a normal computer. The image processing apparatus is mainly configured by a CPU 100 that controls the whole. The CPU 100 is connected via a bus to a ROM 102 for storing programs and the like and a RAM 104 for storing data and the like, and uses a flexible disk (FD) 106, a hard disk (HD) 110 and a CD-ROM 114 as recording media. It is connected to the flexible disk device 108, the hard disk device 112, and the CD-ROM device 116. Further, the CPU 110 is connected to a display device 108 for displaying various screens and the like, a keyboard 120 and a mouse 122 as input means for performing various inputs and instruction operations, and a scanner 124 for reading images. Further, it can be connected to an external digital camera 128, a video camera (not shown), or the like via a USB interface 126. Here, in addition to the operating system program, applications such as an image processing program described later are stored in the hard disk 110. A color face image as a digital image is input from, for example, a scanner 124, a digital camera 128, or the like. Also, the information may be received from another computer or the like from the communication device 130 through a network. The image processing program can be stored in a recording medium such as the flexible disk 106 and the CD-ROM 114. Further, an image processing program recorded on a recording medium such as the flexible disk 106 and the CD-ROM 114 can be downloaded to the image processing apparatus. The configuration of this image processing apparatus is the same in the image processing apparatuses of other embodiments described below, and a description thereof will be omitted in the following embodiments. Further, the image processing described below can be applied to an image processing apparatus in a printer, a digital camera, or the like, and can also be applied to image processing in a mobile phone or the like. In these, the form of the image input is different, but the processing on the input image can be performed similarly.
[0014]
FIG. 5 is a functional block diagram of an area extraction process in the image processing apparatus shown in FIG. The image input unit 200 inputs face image data. In the initial point setting unit 202, the operator sets an arbitrary point in the face image 10 as the initial point 14 for the face image 10 displayed on the screen of the display device 108. In the case of a portrait or the like, since the face is at the center of the face image, the initial point 14 is set, for example, near the center of the face image (see FIG. 1). Next, in the extraction straight line setting unit 204, the operator sets a plurality of straight lines 16 passing through the initial point 14. The pixel feature calculation unit 206 calculates the feature amount of the pixel on this straight line. The feature amount of a pixel is a skin color likeness of a pixel color. For example, the feature value is a value indicating whether or not the R, G, B values of the pixel exist in the skin color definition area. In the pixel selection unit 208, pixels having skin color characteristics are selected based on the feature amounts calculated by the pixel characteristic calculation unit 206, whereby the boundary points 18 between non-skin colors and skin colors and the boundaries between skin colors and non-skin colors are selected. Point 18 is extracted. For example, a threshold value N of the number of consecutive pixels of skin color likeness is determined, and points where the skin color likeness continues N or more are extracted as boundary points 18 from non-skin color to skin color, and the skin color likeness is continuously reduced to less than N. The resulting point is extracted as a boundary point 18 from the skin color to the non-skin color. Further, an edge from skin color to non-skin color may be detected as an extraction point 18 on each straight line 16.
[0015]
Next, based on the boundary points 18 obtained by the pixel selection unit 208, the extraction parameter curve setting unit 210 calculates appropriate parameter group values that describe the shape passing through those boundary points 18. Here, equation (1)
(Equation 1)

Is set as an extraction curve, and the parameter calculation unit 212 uses the coordinate values of the pixels of the four boundary points 18 determined by the pixel selection unit 206 to pass through those points 18. Is solved, and the values of the parameters a, b, x0, and y0 in the equation (1) are obtained by numerical analysis. In this way, an area that can be described by the elliptic curve of the equation (1) is extracted from the image. When an appropriate parameter group is obtained, the parameter curve output unit 214 outputs the elliptic curve 20 described by the parameter value.
[0016]
The number of extracted straight lines is set in the extracted straight line setting unit 204 in advance. For example, if the number of extracted straight lines is set to 3, the extracted straight line setting unit 204 sets three extracted straight lines, and the pixel feature calculating unit 206 calculates a feature amount for these three extracted straight lines. The pixel selection unit 208 obtains the boundary point 18 between the skin color and the non-skin color for the three extracted straight lines. The parameter calculator 212 calculates a parameter value describing the elliptic curve based on the result.
[0017]
FIG. 6 is a flowchart of the area extraction by the CPU 100. First, a face image is input (S100). Next, an initial point 14 set by the operator is input to the input face image (S102), and a plurality of (for example, two) straight lines 16 passing through the initial point 14 set by the operator are input. (S104). Next, the feature amounts of the pixels on these straight lines are calculated (S106), and the pixels having the flesh color feature are selected based on the calculated feature amounts. Then, a boundary point 18 between the non-skin color and the flesh color is obtained along the straight line 16 (S108). Next, parameter values of a mathematical expression (for example, Expression (1)) representing an appropriate shape (for example, an elliptic curve) passing through the boundary point 18 are calculated by numerical analysis (S110). In this way, an area that can be described by the elliptic curve of the equation (1) is extracted from the image. Then, the elliptic curve 20 described by those parameter values is output (S112).
[0018]
Embodiment 2. FIG.
FIG. 7 is a diagram for explaining an area extraction process according to the second embodiment. Here, the validity of a parameter group (first curve parameter group) describing the elliptic curve obtained by the method of the first embodiment is evaluated. For this reason, a second point (referred to as an evaluation point) 24 different from the initial point 14 is set inside the area described by the obtained curve parameters, and the second point 24 passing through the second point 24 is set as in the case of the initial point 14. The straight line 26 is set, and an appropriate curve description parameter group (second curve parameter group) is calculated. Then, the two are compared to evaluate the validity of the curve parameter group. Since a parameter value is obtained not only from one point (initial point 14) but also a second point (evaluation point) 24 is prepared and the parameter value obtained using the initial point 14 is evaluated, high accuracy is obtained. Parameter calculation becomes possible.
[0019]
FIG. 8 is a functional block diagram of an area extraction process in the image processing apparatus shown in FIG. When extracting a region that can be described by a certain parameter curve from an image, first, as in the first embodiment, two or more lines 16 passing through an arbitrary point (initial point 14) in the image are set, and A pixel (boundary point 18) at the boundary between the skin color and the non-skin color is selected based on the feature amount of the pixel (pixel), and an appropriate ellipse parameter is calculated based on the selected pixel group.
[0020]
Next, in the evaluation point setting unit 216, the operator sets one point other than the initial point 14 inside the elliptic curve 20 obtained by the parameter calculation unit 212 as the evaluation point 24 (see FIG. 7), and In the setting unit 204, the operator sets two straight lines 26 passing through the evaluation points 24. Hereinafter, as in the case of the straight line passing through the initial point 14, the pixel feature calculating unit 206 and the pixel selecting unit 208 determine a specific feature based on the feature amount of a pixel (pixel) on two straight lines passing through the evaluation point. A pixel (boundary point 28) having the selected pixel is selected. Next, the parameter calculation unit 212 obtains a parameter group describing an elliptic curve passing through the four boundary points 28. Note that the number of extraction curves passing through the evaluation points 24 may be increased to three or more as in the case of the extraction straight line passing through the initial point.
[0021]
The parameter evaluation unit 220 calculates a least square error between the value of the parameter group obtained using the initial point 14 and the value of the parameter group obtained using the evaluation point 24, and sets the parameter set in the parameter threshold setting unit 218. Compare with threshold. The parameter curve output unit 214 outputs the elliptic curve 20 described by the parameter group obtained by using the initial point 14 when the least square error is evaluated to be equal to or less than the parameter threshold.
[0022]
FIG. 9 is a flowchart of image processing according to the second embodiment. First, a face image is input (S100). Next, an initial point 14 set by the operator is input to the input face image (S102), and a plurality of (for example, two) straight lines 16 passing through the initial point 14 set by the operator are input. (S104). Next, the feature amounts of the pixels on these straight lines are calculated (S106), and the pixels having the flesh color feature are selected based on the calculated feature amounts. Then, a boundary point 18 between the non-skin color and the skin color is obtained (S108). Next, parameter values of a mathematical expression (for example, Expression (1)) representing an appropriate shape (for example, an elliptic curve) passing through the boundary point 18 are calculated by numerical analysis (S110). In this way, an area that can be described by the elliptic curve of the equation (1) is extracted from the image. The above processing is the same as in the first embodiment.
[0023]
Next, one point other than the initial point 14 set by the operator within the obtained elliptic curve 20 is input as the evaluation point 24 (S112), and a plurality of points passing through the evaluation point 24 set by the operator are input. A straight line 26 is set (S114). Hereinafter, similarly to the case of the straight line passing through the initial point 14, the feature amounts of the pixels on these straight lines 26 are calculated (S116), and the pixels having the skin color features are selected based on the calculated feature amounts, and the straight line is selected. A boundary point 28 between the non-skin color and the flesh color is determined along the line 26 (S118). Next, the values of the curve parameters are determined based on the four extracted points 28 (S120). Then, for the evaluation of the curve parameters, the least square error between the parameter value obtained by using the initial point 14 and the parameter value obtained by using the evaluation point 24 is obtained and compared with the parameter threshold (S122). If the least square error is evaluated to be equal to or less than the parameter threshold, an elliptic curve described by the parameter value obtained at the initial point 14 is output (S124).
[0024]
Embodiment 3 FIG.
An image may include multiple human faces. In such a case, as described below, the image is divided into a plurality of regions, and elliptic regions are respectively extracted. Thus, the detection of the faces of the two persons is performed at high speed with a small amount of calculation processing.
[0025]
FIG. 10 shows a case where the skin color regions 42 and 44 of two faces are present separately in one face image 40. First, similarly to the image processing of the second embodiment, the operator sets an initial point 46 and sets two extraction straight lines 48 passing through the initial point 46. The initial point 44 is set on the boundary between the two skin color regions 42 and 44. Then, the characteristic amounts of the pixels (pixels) on these straight lines are obtained, and the pixels having the specific characteristics (sleeve point 50) are selected. Then, a parameter describing an appropriate elliptic curve is calculated based on the sleeve protruding point 50, and an elliptic curve 52 is obtained.
[0026]
Next, for evaluation of region division, the operator sets an evaluation point 54 at a position different from the initial point 46 in the elliptic curve 52, and sets two extraction straight lines 56 passing through the evaluation point 54. Next, based on the feature amounts of the pixels (pixels) on these straight lines, a pixel (sleeve point 58) having a specific feature is selected. Then, a parameter value describing the elliptic curve is calculated from the sleeve appearance point 58. The elliptic curve 60 indicates the ellipse thus obtained.
[0027]
Next, a least square error between the parameter value obtained by using the initial point 46 and the parameter value obtained by using the evaluation point 54 is obtained, and the parameter value is evaluated by comparing with a region division threshold value. When the area division threshold value is exceeded, as shown in FIG. Then, for each initial point 62, two straight lines 64 passing through the initial point 62 at the time of division are set. Similarly to the case of the initial point 46, although not shown, extraction points are obtained along the straight line, and these extraction points are obtained. The elliptical area is obtained based on
[0028]
FIG. 12 shows a case where the skin color regions 72 and 74 of two faces overlap with one face image 70. First, similarly to the image processing of the second embodiment, the operator sets an initial point 76 and sets a plurality of extraction straight lines 78 passing through the initial point 76. The initial point 44 is set on the boundary between the two skin color regions 72 and 74. Then, based on the feature amount of the pixel (pixel) on the straight line 78, a pixel (sleeve point 80) having a specific feature is selected. Then, a parameter value describing an appropriate elliptic curve is calculated from the sleeve appearance point 80, and an elliptic curve 82 is obtained.
[0029]
Next, in order to evaluate the area division, the operator sets an evaluation point 84 in the elliptic curve 82 at a position different from the initial point 76, and sets two extraction straight lines 86 passing through the evaluation point 84. Next, a pixel (extraction point 88) having a specific feature is selected based on the feature amount of the pixel on the straight line 86. Then, from the extraction points 88, a parameter value describing an appropriate elliptic curve is calculated. The elliptic curve 90 shows the ellipse thus obtained.
[0030]
Next, the least square error between the parameter obtained by using the initial point 76 and the parameter obtained by using the evaluation point 84 is obtained, and the elliptic parameter is evaluated by comparing with the area division threshold. When the threshold value exceeds the region division threshold, division initial points 92 are set on the left and right of the initial point 76 as shown in FIG. The position of the initial point at the time of division is set to the left side of the skin color area 72 and the right side of the skin color area 74, and the straight line 94 is set so as not to extend over the two skin color areas. Then, two straight lines 94 passing through the initial point 92 at the time of division are set for each initial point 92 at the time of division, and, similarly to the initial point 76, extraction points are obtained along a straight line (not shown), The elliptical area is obtained based on
[0031]
FIG. 14 is a functional block diagram of image processing in the image processing apparatus shown in FIG. When extracting a region that can be described by a certain parameter curve from an image, first, similarly to the first embodiment, in blocks 200 to 212, two or more lines passing through arbitrary points (initial points) in the image A pixel having a specific feature is selected based on the feature amount of the pixel (pixel), a boundary between the skin color and the non-skin color is obtained, and an appropriate parameter value is calculated from the selected pixel group.
[0032]
When the parameter evaluation unit 220 evaluates that the value exceeds the region division threshold set by the region division threshold setting unit 222 (for example, in the case of FIGS. 10 and 11), the division initial point setting unit 224 As shown in the example of FIG. 12 and FIG. 13, the initial points 62 and 92 at the time of division are set to the left and right of the initial points 40 and 70, and the same calculation as that of the first embodiment is performed for each of the initial points 62 and 92 at the time of division. To extract an elliptical area.
[0033]
FIG. 15 is a flowchart of image processing according to the third embodiment. First, a face image is input (S100). Next, initial points 46 and 76 set by the operator are input to the input face image (S102), and a plurality of (for example, two) passing through the initial points 46 and 76 set by the operator. The straight lines 48 and 78 are input (S104). Next, the feature amounts of the pixels on these straight lines are calculated (S106), and the pixels having the flesh color feature are selected based on the calculated feature amounts. Then, the boundary points 50 and 80 between the non-skin color and the skin color are obtained (S108). Next, parameter values of a mathematical expression (for example, Expression (1)) representing an appropriate curve (ellipse) passing through the boundary points 50 and 80 are calculated by numerical analysis (S110). In this way, an area that can be described by the elliptic curve of the equation (1) is extracted from the image.
[0034]
Next, one point other than the initial points 46 and 76 set by the operator in the obtained elliptic curve is input as evaluation points 54 and 84 (S112), and the evaluation points 54 and 84 set by the operator are input. A plurality of straight lines 56 and 86 passing through 84 are set (S114). Hereinafter, similarly to the case of the straight lines 48 and 78 passing through the initial points 46 and 76, the feature amounts of the pixels on these straight lines are calculated (S116), and the pixels having the skin color features are selected based on the calculated feature amounts. I do. Then, the boundary points 50 and 80 between the non-skin color and the skin color are obtained (S118). Next, a parameter value is determined based on the four boundary points 50 and 80 (S120). Then, for parameter value evaluation, the least square error between the parameter value obtained using the initial points 46 and 76 and the parameter value obtained using the evaluation points 54 and 84 is obtained and compared with the region division threshold value. (S126).
[0035]
If it is evaluated that the least square error exceeds the region division threshold value, division initial points 62 and 92 are set to the left and right of the initial points 46 and 76 (S128), and a plurality of initial points 62 and 92 passing through the division initial points 62 and 92 are set. The straight lines 64 and 94 are set (S130). Next, the feature amounts of the pixels on these straight lines 64 and 94 are calculated (S132), and a pixel having a flesh color feature is selected based on the calculated feature amounts. Then, a boundary point between the non-skin color and the flesh color is obtained along the straight lines 64 and 94 (S134). Next, a parameter value of a mathematical expression (for example, Expression (1)) representing an appropriate curve (ellipse) passing through the boundary point is calculated by numerical analysis (S136). In this way, a plurality of regions that can be described by the elliptic curve of Expression (1) are extracted from the image.
[0036]
In the above embodiment, the initial point, the evaluation point, and the initial point at the time of division are selected from the same single image. For example, when the input image is a moving image, it is possible to select from different frames. is there.
[0037]
【The invention's effect】
Extraction of a region from an image and extraction of parameter values of a mathematical expression describing the shape of the region can be performed with a small amount of calculation processing.
Individual extraction and individual parameter value extraction of a plurality of overlapping regions in an image can be performed with a small amount of calculation processing.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating image processing according to the first embodiment; FIG. 2 is a diagram illustrating an elliptical region obtained by the image processing described with reference to FIG. 1; FIG. 3 is a diagram illustrating a modification of the image processing according to the first embodiment; FIG. 4 is a block diagram of an image processing apparatus. FIG. 5 is a functional block diagram of information processing according to the first embodiment. FIG. 6 is a flowchart of information processing according to the first embodiment. FIG. 8 is a functional block diagram of information processing according to the second embodiment. FIG. 9 is a flowchart of information processing according to the second embodiment. FIG. 10 is a diagram illustrating a first example of image processing according to the third embodiment. FIG. 11 is a diagram illustrating another example of image processing according to the third embodiment. FIG. 12 is a diagram illustrating a second example of area division processing according to the third embodiment. The area division processing of the second example in mode 3 Functional block diagram of an information processing of another Figure 14] Embodiment 3 of the bright FIG. 15 is a flowchart of processing according to the third embodiment EXPLANATION OF REFERENCE NUMERALS
10 images, 12 skin color areas, 14 initial points, 16 straight lines, 18 boundary points, 20 elliptic curves, 24 evaluation points, 26 straight lines, 46 initial points, 48 straight lines, 50 boundary points, 54 evaluation points, 56 straight lines, 58 boundary points , 62 initial points at the time of division, 64 straight lines, 76 initial points, 78 straight lines, 80 boundary points, 84 evaluation points, 86 straight lines, 88 boundary points, 92 initial points at the time of division, 94 straight lines, 100 CPU, 102 ROM, 106 flexible disk , 108 display device, 110 hard disk, 114 CD-ROM, 120 keyboard, 122 mouse.

Claims (9)

An image processing apparatus for extracting an area surrounded by a shape represented by a mathematical expression including a parameter from an image,
Image input means for inputting an image,
Initial point setting means for setting one point in the input image as an initial point;
Straight line setting means for setting a plurality of straight lines passing through the initial point,
For pixels on the plurality of straight lines, based on pixel data, a boundary detecting means for detecting a boundary point of an area where pixels having specific characteristics are continuously present,
Parameter calculation means for calculating the value of the parameter of the formula so as to pass through the boundary point,
An output unit for outputting a shape described by the parameter value. further,
Evaluation point setting means for setting a second point different from the initial point inside the shape described by the parameter value as an evaluation point;
Second straight line setting means for setting a second plurality of straight lines passing through the evaluation point;
A second boundary detection unit that detects, based on the pixel data, a second boundary point of an area where pixels having specific characteristics are continuously present, for pixels on the second plurality of straight lines;
A second parameter calculating means for calculating a second value of the parameter of the mathematical expression so as to pass through the second boundary point, a parameter value obtained by the parameter calculating means, and a parameter value obtained by the second parameter calculating means. Evaluation means for evaluating whether or not the difference is within a predetermined range,
2. The image processing apparatus according to claim 1, wherein the output unit outputs a shape described by the parameter value when the difference is within a predetermined range. An image processing apparatus for extracting an area surrounded by a shape represented by a mathematical expression including a parameter from an image,
Image input means for inputting an image,
Initial point setting means for setting one point in the input image as an initial point;
First straight line setting means for setting a first plurality of straight lines passing through the initial point;
First boundary detection means for detecting, based on pixel data, a first boundary point of a region where pixels having specific characteristics are continuously present, for pixels on the plurality of straight lines;
First parameter calculating means for calculating the value of the parameter of the mathematical formula so as to pass through the boundary point;
Evaluation point setting means for setting a second point different from the initial point inside the shape described by the parameter value as an evaluation point;
Second straight line setting means for setting a second plurality of straight lines passing through the evaluation point;
A second boundary detecting means for detecting, based on the pixel data, a second boundary point of a region where pixels having specific characteristics are continuously present, for pixels on the second plurality of straight lines; ,
A second parameter calculating means for calculating a second value of the parameter of the mathematical expression so as to pass through the second boundary point; a parameter obtained by the first parameter calculating means; and a parameter obtained by the second parameter calculating means. Evaluation means for evaluating whether the value difference is within a predetermined range,
If the difference exceeds a predetermined range, two second initial points are set on different sides with respect to the initial point;
Third straight line setting means for setting a plurality of straight lines passing through each of the second initial points;
A third boundary detection unit that detects, based on the pixel data, a third boundary point of a region where pixels having specific characteristics are continuously present, for pixels on the plurality of straight lines;
Third parameter calculating means for calculating a third value of a parameter describing the mathematical expression so as to pass through the third boundary point;
An output unit that outputs each shape described by the third parameter value. An image processing method for extracting an area surrounded by a shape represented by a mathematical expression including a parameter from an image,
Enter the image,
Set one point in the input image as the initial point,
Set a plurality of straight lines passing through the initial point,
For the pixels on the plurality of straight lines, based on the pixel data, to detect a boundary point of an area where pixels having specific characteristics are continuously present,
Calculating the values of the parameters of the formula so as to pass through the boundary point,
An image processing method for outputting a shape described by the parameter value. further,
A second point different from the initial point inside the shape described by the parameter value is set as an evaluation point,
Setting a second plurality of straight lines passing through the evaluation points;
For pixels on the second plurality of straight lines, based on the pixel data, a second boundary point of a region where pixels having specific characteristics are continuously present is detected,
A second value of the parameter of the mathematical expression is calculated so as to pass through the second boundary point, and a difference between the parameter value obtained based on the boundary point and the parameter value obtained based on the second boundary point is within a predetermined range. Evaluate whether there is
5. The image processing method according to claim 4, wherein when the difference is within a predetermined range, a shape described by the parameter value is output. An image processing method for extracting an area surrounded by a shape represented by a mathematical expression including a parameter from an image,
Enter the image,
Set one point in the input image as the initial point,
Setting a first plurality of straight lines passing through the initial point;
For pixels on the plurality of straight lines, based on the pixel data, detect a first boundary point of a region where pixels having specific characteristics are continuously present,
Calculating a first value of a parameter describing the mathematical expression so as to pass through the boundary point;
A second point different from the initial point inside the shape described by the parameter value is set as an evaluation point,
Setting a second plurality of straight lines passing through the evaluation points;
For pixels on the second plurality of straight lines, based on the pixel data, a second boundary point of a region where pixels having specific characteristics are continuously present is detected,
Calculating a second value of the parameter to pass through the second boundary point;
Evaluate whether or not the difference between the parameter values determined based on the boundary point and the second boundary point is within a predetermined range,
If the difference exceeds a predetermined range, two second initial points are set on different sides with respect to the initial point;
Setting a plurality of straight lines passing through each of the second initial points;
For pixels on the plurality of straight lines, a third boundary point of an area where pixels having specific characteristics are continuously present is detected based on the pixel data,
Calculating a third value of the parameter to pass through the third boundary point;
An image processing method for outputting each shape described by a third value of the parameter. An image processing program for extracting an area surrounded by a shape represented by a mathematical expression including a parameter from an image,
Inputting an image;
Setting one point in the input image as an initial point;
Setting a plurality of straight lines passing through the initial point,
For pixels on the plurality of straight lines, based on the pixel data, a step of detecting a boundary point of an area where pixels having specific characteristics are continuously present,
Calculating a value of a parameter describing the mathematical formula so as to pass through the boundary point;
An image processing program executed by a computer, comprising a step of outputting a shape described by the parameters. further,
Setting a second point different from the initial point as an evaluation point inside the shape described by the parameter value;
Setting a second plurality of straight lines passing through the evaluation points;
Detecting, for pixels on the second plurality of straight lines, a second boundary point of a region where pixels having specific characteristics are continuously present, based on the pixel data;
Calculating a second value of a parameter describing the mathematical expression so as to pass through the second boundary point;
Evaluating whether the difference between the parameter values obtained based on the boundary point and the second boundary point is within a predetermined range,
8. The method according to claim 7, wherein in the step of outputting the shape described by the parameter value, when the difference is within a predetermined range, the shape described by the parameter value is output. Image processing program. An image processing program for extracting an area surrounded by a shape represented by a mathematical expression including a parameter from an image,
Inputting an image;
Setting one point in the input image as an initial point;
Setting a first plurality of straight lines passing through the initial point;
Detecting, for pixels on the plurality of straight lines, a first boundary point of a region where pixels having specific characteristics are continuously present, based on the pixel data;
Calculating a first value of a parameter describing the mathematical expression so as to pass through the boundary point;
Setting a second point different from the initial point inside the shape described by the parameter value as an evaluation point;
Setting a second plurality of straight lines passing through the evaluation points;
Detecting, for pixels on the second plurality of straight lines, a second boundary point of a region where pixels having specific characteristics are continuously present, based on the pixel data;
Calculating a second value of the parameter to pass through the second boundary point;
Evaluating whether or not the difference between the parameter values determined based on the boundary point and the second boundary point is within a predetermined range,
Setting two second initial points on different sides with respect to the initial point if the difference exceeds a predetermined range;
Setting a plurality of straight lines passing through each of the second initial points;
Detecting, for pixels on the plurality of straight lines, a third boundary point of an area where pixels having specific characteristics are continuously present, based on the pixel data;
Calculating a third value of the parameter to pass through the third boundary point;
Outputting each shape described by the third value of the parameter, and an image processing program executed by a computer.

Images