Method of extracting feature image data and method of extracting person's face data

A method of extracting feature image data, comprising the steps of:

• (a) dividing a color original image into a plurality of picture elements and resolving each of said picture elements into red, green, and blue colors, wherein each color is photometrically measured;
• (b) obtaining a two-dimensional histogram of hue and saturation values in accordance with the data obtained through the photometry;
• (c) cutting the two-dimensional histogram by a feature plane parallel with a coordinate plane, including a cross section corresponding to the crest of a certain mountain of the obtained two-dimensional histogram, and also including hue- and saturation-value axes of the two-dimensional histogram to obtain the cross sections of the two-dimensional histograms;
• (d) expanding each cross section until the interval between boundaries of adjacent cross sections on the feature plane decreases in a predetermined value or less, or performing exclusive expansion for expanding cross sections until the frequency of cross-sectional boundaries decreases to a minimum value or a predetermined value close to the minimum value or less, to divide every mountain of the two-dimensional histogram at cross-sectional boundaries after exclusive expansion;
• (e) determining which divided mountain each picture element of the color original image belongs to in order to divide picture elements into groups corresponding to the respective divided mountains and dividing the color original image into a plurality of divided regions corresponding to said groups;
• and
• (f) selecting at least one of the divided regions to extract the data for the selected region as feature image data.

A method of extracting feature image data, comprising the steps of:

• (a) dividing a color original image into a plurality of picture elements and resolving each picture element into red, green, and blue colors, wherein each of said colors is photometrically measured;
• (b) obtaining a two-dimensional histogram of hue and saturation values in accordance with data obtained through the photometry and dividing the histogram into mountains;
• (c) dividing each mountain of the obtained two-dimensional histogram into a plurality of small mountains each with one crest, respectively, by a feature plane which is parallel with a coordinate plane including hue- and saturation-value axes of the two-dimensional histogram and includes the largest cross section corresponding to only the crest of one of said small mountains;
• (d) determining which of said small mountains each picture element of the color original image is included in to classify the picture elements into groups corresponding to each of said small mountains;
• (e) obtaining degree of similarity between at least one picture element included in each of said small mountains and picture elements excluded from said small mountains for each small mountain in accordance with hue and saturation values;
• (f) unifying the picture elements excluded from said small mountain into unified groups corresponding to each small mountain with a high degree of similarity;
• (g) dividing the color original image into a plurality of divided regions corresponding to each of said unified groups;
• and
• (h) selecting at least one of the divided regions to extract the data for the selected region as feature image data.

A method of extracting human face image data, comprising the steps of:

• (a) dividing a color original image into a plurality of picture elements and resolving each picture element into red, green, and blue colors, wherein each of said colors is photometrically measured;
• (b) obtaining color regions with the same or similar hue on said color original image in accordance with data obtained through the photometry;
• (c) assuming circles or ellipses inscribed with respect to boundary portions of the obtained color regions and dividing the color regions obtained using circular or elliptic regions starting from the largest circle or ellipse in order to thereby obtain a plurality of divided regions;
• and
• (d) selecting at least one of the divided regions from which to extract the data of the selected region as human face image data.

A method of extracting human face image data, comprising the steps of:

• (a) dividing a color original image into a plurality of picture elements and resolving each picture element into three colors of red, green, and blue, wherein each of said colors is photometrically measured;
• (b) obtaining color regions with the same or similar hue on said color original image in accordance with data obtained through the photometry;
• (c) assuming circles or ellipses inscribed with respect to boundary portions of the obtained color regions to thereby obtain the largest circular or elliptic region;
• (d) obtaining a convex region having a predetermined or larger area protruding outward from the circumference of a circle or ellipse and contacted to said obtained region to thereby divide said color region by the circular region inscribed with respect to the boundary portion of the color regions included in said convex portion to thereby obtain a plurality of divided regions;
• and
• (e) selecting at least one of the divided regions from which to extract the data of the selected region as human face image data.