Image segmentation method and image segmentation device
View Patent ↗An image segmentation method and an image segmentation device are provided. The method comprises receiving a video image of a dynamic movement of a target object, acquiring a full-image optical flow of the video image to estimate a first displacement of each pixel therein, acquiring a background optical flow of the video image to estimate a second displacement of a background pixel therein; comparing the first displacement with the second displacement to obtain a foreground region of the target object; extracting feature points in the video image in the foreground region, calculating a probability density of the feature points to determine a number of the target objects; performing visual tracking and movement trajectory analysis on the target object to track the same; performing stationary judgment and image segmentation on the target object according to an interframe displacement of the feature points, an interframe cutting window similarity and tracking box scaling.
1. A computer-implemented image segmentation method, comprising:
receiving, at one or more computing devices, a video image of a dynamic movement of a target object;
acquiring, at the one or more computing devices, a full-image optical flow of the video image and a background optical flow of the video image;
estimating, at the one or more computing devices, a first displacement of each pixel in the video image based on the acquired full-image optical flow of the video image;
estimating, at the one or more computing devices, a second displacement of a background pixel in the video image based on the acquired background optical flow of the video image;
comparing, at the one or more computing devices, the first displacement with the second displacement to obtain a foreground region of the target object;
extracting, at the one or more computing devices, feature points in the video image in the foreground region of the target object;
calculating, at the one or more computing devices, a probability density of the feature points extracted to determine a number of the target objects;
performing, at the one or more computing devices, a visual tracking and a movement trajectory analysis on the target object to track the target object; and
performing, at the one or more computing devices, a stationary judgment and an image segmentation on the target object according to an interframe displacement of the feature points, an interframe cutting window similarity, and a tracking box scaling.
2. The image segmentation method according to claim 1 , wherein receiving the video image of the dynamic movement of the target object comprises:
receiving, at the one or more computing devices, the video image of the dynamic movement of the target object input via a camera.
3. The image segmentation method according to claim 1 , further comprising:
receiving at the one or more computing devices, an image identification request input via speech, a touch screen, or a dynamic sensor so as to trigger an operation of receiving the video image.
4. A computer-implemented image segmentation device, comprising:
one or more processors configured to execute one or more software modules, the one or more software modules including:
a receiving module configured to receive a video image of a dynamic movement of a target object;
an acquiring module configured to acquire a full-image optical flow of the video image to estimate a first displacement of each pixel in the video image, and to acquire a background optical flow of the video image to estimate a second displacement of a background pixel in the video image;
a comparing module configured to compare the first displacement with the second displacement to obtain a foreground region of the target object;
an extracting module configured to extract feature points in the video image in the foreground region of the target object;
a calculating module configured to calculate a probability density of the feature points extracted to determine a number of the target objects;
a tracking module configured to perform a visual tracking and a movement trajectory analysis on the target object to track the target object; and
a segmenting module configured to perform a stationary judgment and an image segmentation on the target object according to an interframe displacement of the feature points, an interframe cutting window similarity and a tracking box scaling.
5. The image segmentation device according to claim 4 , further comprising:
a camera, wherein the receiving module is configured to receive the video image of the dynamic movement of the target object input via the camera.
6. The image segmentation device according to claim 4 , wherein the receiving module is further configured to receive an image identification request input via speech, a touch screen or a dynamic sensor so as to trigger an operation of receiving the video image.
7. A non-transitory computer readable storage medium, comprising a computer program configured to execute the following steps on one or more computing devices:
receiving a video image of a dynamic movement of a target object, acquiring a full-image optical flow of the video image to estimate a first displacement of each pixel in the video image, and acquiring a background optical flow of the video image to estimate a second displacement of a background pixel in the video image;
comparing the first displacement with the second displacement to obtain a foreground region of the target object;
extracting feature points in the video image in the foreground region of the target object, and calculating a probability density of the feature points extracted to determine a number of the target objects;
performing a visual tracking and a movement trajectory analysis on the target object to track the target object; and
performing a stationary judgment and an image segmentation on the target object according to an interframe displacement of the feature points, an interframe cutting window similarity and a tracking box scaling.