Face tracking method for electronic camera device

1. A face tracking method for electronic camera devices, which is applied to an electronic camera device having a face database and a face classifier therein, wherein the face database is provided for storing data including face position, size and skin color prototype of a previously stored preview image, comprising the steps of:

obtaining a current preview image;

determining whether or not a known face exists in the face database;

defining a searching space on the current preview image;

using the face classifier to detect the searching space in the current preview image, and determining whether or not a face exists in the searching space;

obtaining a skin color extraction information to define the searching space, when the face database is determined to have a known face, according to the data including a position, a size and a skin color prototype of the known face in the face database;

automatically updating the corresponding face data in the face database, when the data of each face in the current preview image is detected;

starting a face tracking mechanism to extend the searching space into an extended searching region according to the data including the position, size and skin color prototype of a know face in the face database, when the face classifier cannot detect a face in the searching space; and

calculating a color histogram of each position in the extended searching region and a color histogram matching of a corresponding known face, and using a data of a corresponding position with the maximum histogram matching for the detection as the detected data of each face in the current preview image.

2. The method of claim 1 , further comprising the step of:

stabilizing a face rectangle that labels the face position and size, when the data of each face in the current preview image is detected.

3. The method of claim 2 , wherein the stabilization process comprises the steps of:

reading the position and size information of a face rectangle in a plurality of recent preview images from the face database;

performing a median filtering to the position and size information of the face rectangle in the preview images;

perform an exponential smooth process to a median filtering result and a history data of the position and size of the known face in the face database; and labeling the median filtered and exponential smooth processed face rectangle in the corresponding face position.

4. The method of claim 3 , wherein the skin color prototype is represented by a color histogram information, and the color histogram information is a color statistic value of the face position and size on the preview image.

5. A face tracking method for electronic camera devices, which is applied to an electronic camera device having a face database and a face classifier therein, wherein the face database is provided for storing data including face position, size and skin color prototype of a previously stored preview image, comprising the steps of:

obtaining a current preview image;

determining whether or not a known face exists in the face database;

defining a searching space on the current preview image;

using the face classifier to detect the searching space in the current preview image, and determining whether or not a face exists in the searching space;

defining the searching space according to the position and size of each skin color extraction area in the current preview image, when it is determined that there is no known face in the face database;

automatically updating the corresponding face data in the face database, when a data of each face in current preview image is detected; and

stabilizing a face rectangle that labels the face position and size, when the data of each face in the current preview image is detected, wherein the stabilization process comprises the steps of:

reading the position and size information of a face rectangle in a plurality of recent preview images from the face database;

performing a median filtering to the position and size information of the face rectangle in the preview images;

perform an exponential smooth process to a median filtering result and a history data of the position and size of the known face in the face database; and

labeling the median filtered and exponential smooth processed face rectangle in the corresponding face position.

6. The method of claim 5 , wherein the skin color prototype is represented by a color histogram information, and the color histogram information is a color statistic value of the face position and size on the preview image.

7. A face tracking method for electronic camera devices, which is applied to an electronic camera device having a face database and a face classifier therein, wherein the face database is provided for storing data including face position, size and skin color prototype of a previously stored preview image, comprising the steps of:

obtaining a current preview image;

determining whether or not a known face exists in the face database;

defining a searching space on the current preview image;

using the face classifier to detect the searching space in the current preview image, and determining whether or not a face exists in the searching space; and

stabilizing a face rectangle that labels the face position and size, when the data of each face in the current preview image is detected, wherein the stabilization process comprises the steps of:

reading the position and size information of a face rectangle in a plurality of recent preview images from the face database;

performing a median filtering to the position and size information of the face rectangle in the preview images;

perform an exponential smooth process to a median filtering result and a history data of the position and size of the known face in the face database; and

labeling the median filtered and exponential smooth processed face rectangle in the corresponding face position.

8. The method of claim 7 , wherein the skin color prototype is represented by a color histogram information, and the color histogram information is a color statistic value of the face position and size on the preview image.