Image processing apparatus, image processing method, and interface apparatus
In order to detect a specific detection object from an input image, a color serving as a reference is calculated in a reference image region. The difference for each color component between each pixel in the detection window and the reference color is calculated. Whether or not the detection object is included in the detection window is discriminated by a feature vector indicating how the difference is distributed in the detection window.
1. An image processing apparatus for detecting an object to be detected in an input image, comprising:
an extraction unit configured to extract a reference image region from the input image;
a reference color calculation unit configured to calculate a reference color from the reference image region;
a feature vector calculation unit configured to (a) calculate color differences between color components of each pixel in a detection window being set in the input image and color components of the reference color, (b) calculate a histogram regarding the color differences in each of local regions, the local regions being defined by dividing an inside of the detection window, and (c) calculate a feature vector including the calculated histograms as elements of the feature vector, each of the elements corresponding to the position of each local region in the detection window; and
a discrimination unit configured to discriminate whether or not the object is included in the detection window based on the feature vector.
2. The apparatus according to claim 1 , wherein with the feature vector calculation unit calculates each histogram with a quantization width of color being varied-from one bin to another bin in the histogram.
3. The apparatus according to claim 2 , wherein the feature vector calculation unit sets the quantization width narrower within the range of the value of the color component higher in frequency, and setting the quantization width wider in the range of the value of the color component lower in frequency based on color histograms previously calculated from training data of the object.
4. The apparatus according to claim 3 , wherein the feature vector calculation unit, further
(c) determines the gradient direction of the luminance of each local region,
(d) quantizes the gradient direction, and calculates a gradient histogram regarding the direction for each local region, and
(e) combines the histogram with the gradient histogram to obtain a combined histogram for each local region, and calculates the feature vector with the combined histograms as its elements corresponding to the positions of the local regions.
5. An interface apparatus for instructing a command to a target apparatus using a part of an operator, comprising:
an image pickup unit configured to pick up an input image containing the part of the operator;
an extraction unit configured to extract a reference image region to calculate a reference color from the input image;
a reference color calculation unit configured to calculate the reference color from the reference image region;
a feature vector calculation unit configured to (a) calculate a difference for each component in the color space between the color of each pixel in a detection window set in the input image and the reference color, (b) form a histogram regarding the differences in each of local regions obtained defined by dividing a defined by dividing an inside of the detection window, and (c) calculate a feature vector consisting of the calculated histograms as elements of the feature vector, each of the elements corresponding to the position of the local region in the detection window;
a discrimination unit configured to discriminate whether or not the part of the operator is included in the detection window based on the feature vector; and
a control unit configured to give a command to the target apparatus depending upon the discrimination result of the discrimination unit.
6. The apparatus according to claim 5 , wherein the image pickup unit configured to pick up an input image containing a hand or fingers of the operator.
7. An image processing method for detecting an object to be detected in an input image, comprising:
extracting a reference image region from the input image;
calculating a reference color from the reference image region;
calculating color differences between color components of each pixel in a detection window being set in the input image and the color component of the reference color;
calculating a histogram regarding the color differences in each of local regions, the local regions being defined by dividing an inside of the detection window, calculating a feature vector including the calculated histograms as elements of the feature vector, each of the elements corresponding to the position of each local region in the detection window; and
discriminating whether or not the object is included in the detection window based on the feature vector.