Information processing apparatus and image region segmentation method
A captured image acquisition section 50 acquires, from an imaging apparatus 12 , data of a polarized image obtained by capturing a target object and stores the data into an image data storage section 52 . A normal line distribution acquisition section 362 of a region segmentation section 310 acquires a distribution of normal line vectors of a subject surface using the polarized image. A region decision section 312 evaluates the distribution of normal line vectors with a predetermined criterion indicating stability such as a dispersion and determines a region in which the stability is high as a region of a figure of a subject to segment the image. An output data generation section 56 carries out various processes using a result of the segmentation to generate output data.
1. An information processing apparatus, comprising:
a normal line distribution acquisition section configured to acquire a distribution of normal line vectors of a subject surface from a captured polarized image;
a region segmentation section configured to specify a region of a figure for each subject and segment an image plane by evaluating stability of the distribution of the normal line vectors with a given criterion; and
an output data generation section configured to perform a process based on a result of the segmentation to generate output data and output the output data,
wherein the region segmentation section evaluates continuity of a variation with respect to positions of the normal line vectors and specifies a region in which it is decided that there is continuity as the region of the figure.
2. The information processing apparatus according to claim 1 , wherein the region segmentation section evaluates a dispersion of positions of the normal line vectors and specifies a region in which it is decided that the dispersion is small as the region of the figure.
3. The information processing apparatus according to claim 2 , wherein the region segmentation section successively calculates, while the region segmentation section moves a search window of a given size on the image plane, a difference between an average of normal line vectors in the search window and an average of normal line vectors in a reference region set in the search window as an index indicative of the dispersion, and decides a region in which the dispersion is small based on a position of the search window when the index smaller than a given threshold value is obtained.
4. The information processing apparatus according to claim 1 , wherein the region segmentation section evaluates a magnitude of time variation of the normal line vectors and specifies a region in which it is decided that the time variation is small as the region of the figure.
5. The information processing apparatus according to claim 1 , wherein the region segmentation section performs, based on a characteristic of the normal line vectors in the region of the figure specified for each subject, further segmentation in response to an orientation or a shape of a plane that configures the subject.
6. The information processing apparatus according to claim 1 , wherein the region segmentation section detects a position of a normal line vector having a specific direction from within the distribution of the normal line vectors acquired by the normal line distribution acquisition section and searches for a region in which a variation of the normal line vectors has continuity from a base point given by the detected position to specify the region of the figure.
7. The information processing apparatus according to claim 6 , wherein the region segmentation section sets the specific direction based on the normal line vectors in the region of the figure specified in a preceding image frame within the polarized image captured at a given frame rate.
8. The information processing apparatus according to claim 6 , wherein the region segmentation section searches for a region in which the normal line vectors have variation according to a surface shape of a target object prepared in advance to specify the region of the figure.
9. An image region segmentation method by an information processing apparatus, comprising:
acquiring data of a polarized image captured by an imaging apparatus;
acquiring a distribution of normal line vectors of a subject surface from the polarized image;
evaluating stability of the distribution of the normal line vectors with a given criterion to specify a region of a figure for each subject and segmenting an image plane;
evaluating a dispersion of positions of the normal line vectors;
specifying a region in which it is decided that the dispersion is small as the region of the figure;
calculating, while moving a search window of a given size on the image plane, a difference between an average of normal line vectors in the search window and an average of normal line vectors in a reference region set in the search window as an index indicative of the dispersion;
deciding a region in which the dispersion is small based on a position of the search window when the index smaller than a given threshold value is obtained; and
performing processing based on a result of the segmentation to generate output data and outputting the output data to an external apparatus.
10. A non-transitory computer program readable medium having stored thereon a program for a computer, the program comprising:
by a normal line distribution acquisition section, acquiring a distribution of normal line vectors of a subject surface from a captured polarized image;
by a region segmentation section, evaluating stability of the distribution of the normal line vectors with a given criterion to specify a region of a figure for each subject and segmenting an image plane; and
by an output data generation section, performing processing based on a result of the segmentation to generate output data and outputting the output data,
wherein the region segmentation section evaluates a magnitude of time variation of the normal line vectors and specifies a region in which it is decided that the time variation is small as the region of the figure.