Detecting orientation of digital images using face detection information

A method of determining an orientation state of a digital image, comprising:

• using a processor programmed by embedded digital code within a processor-readable medium;
• acquiring a digital image using an image sensor;
• detecting one or more face regions within the acquired digital image, including contemporaneously determining an orientation state of the digital image, including:
• applying a combined face classifier set, that includes one or more rotation invariant classifiers and three or more sets of orientation-specific classifiers, to said digital image, wherein each of the one or more rotation invariant classifiers is common to individual versions of each of the orientation-specific classifiers and included only once in the combined face classifier, and wherein each set of the three or more sets of orientation-specific classifiers is rotated relative to the other sets of orientation-specific classifiers;
• and
• based on results of said applying said combined face classifier set to the digital image, determining an orientation state of the digital image as an orientation among three or more orientations corresponding respectively to the three or more orientation-specific classifier sets with a highest probability of detection of one or more face regions.

One or more non-transitory processor-readable media having code embedded therein for programming the processor to perform a method of determining an orientation state of an acquired digital image, wherein the method comprises:

• detecting one or more face regions within an acquired digital image, including contemporaneously determining an orientation state of the digital image, including:
• applying a combined face classifier set, that includes one or more rotation invariant classifiers and three or more sets of orientation-specific classifiers, to said digital image, wherein each of the one or more rotation invariant classifiers is common to individual versions of each of the orientation-specific classifiers and included only once in the combined face classifier, and wherein each set of the three or more sets of orientation-specific classifiers is rotated relative to the other sets of orientation-specific classifiers;
• and
• based on results of said applying said combined face classifier set to the digital image, determining an orientation state of the digital image as an orientation among three or more orientations corresponding respectively to the three or more orientation-specific classifier sets with a highest probability of detection of one or more face regions.