| Summary: |
With this project it is intended to study, to develop, to implement, to test and to adjust, a set of computational techniques that allow, on one hand, the segmentation of deformable objects (2D/3D) represented in images and, on the other hand, the tracking and the time analysis of the motion (or the
deformation) of the same objects.
The segmentation of objects represented in images is one of the tasks more used in the area of image processing and analysis. For segmentation, it is understood the identification of zones, or areas, of an object, or of a structure, represented in an (2D/3D) image or represented through an (2D/3D) image sequence. Generally, this classification is one of the operations
to be applied first in the resolution of various image analysis problems; for example: in the medical image area, the determination of the left ventricle contour, the identification of arteries or of nodules, etc.; in the area of automatic character recognition, before associating the symbol presented in an image to the respective character, it is necessary to detect it in the same image; in the area of vehicle identification and tracking, the first task is necessarily the determination of the vehicle(s) presented in each image; in the faces image area, generally the first step is the detection of the face(s) presented in each image to consider; etc. The operation of segmentation, besides being an operation applied in static images, it can be applied in image sequences also. Therefore, it is associated to objects motion (or deformation) tracking.
The tracking of objects, or of structures, along an (2D/3D) image sequence,
is one of the most important subjects in the Computational Vision domain, having therefore along the last years a raised attention of researches. The tracking operation consists on the identification of the same object or structure along all the images that compose a temporal sequence. Obviously that it is a complex operation, as generally innumerab  |
Summary
With this project it is intended to study, to develop, to implement, to test and to adjust, a set of computational techniques that allow, on one hand, the segmentation of deformable objects (2D/3D) represented in images and, on the other hand, the tracking and the time analysis of the motion (or the
deformation) of the same objects.
The segmentation of objects represented in images is one of the tasks more used in the area of image processing and analysis. For segmentation, it is understood the identification of zones, or areas, of an object, or of a structure, represented in an (2D/3D) image or represented through an (2D/3D) image sequence. Generally, this classification is one of the operations
to be applied first in the resolution of various image analysis problems; for example: in the medical image area, the determination of the left ventricle contour, the identification of arteries or of nodules, etc.; in the area of automatic character recognition, before associating the symbol presented in an image to the respective character, it is necessary to detect it in the same image; in the area of vehicle identification and tracking, the first task is necessarily the determination of the vehicle(s) presented in each image; in the faces image area, generally the first step is the detection of the face(s) presented in each image to consider; etc. The operation of segmentation, besides being an operation applied in static images, it can be applied in image sequences also. Therefore, it is associated to objects motion (or deformation) tracking.
The tracking of objects, or of structures, along an (2D/3D) image sequence,
is one of the most important subjects in the Computational Vision domain, having therefore along the last years a raised attention of researches. The tracking operation consists on the identification of the same object or structure along all the images that compose a temporal sequence. Obviously that it is a complex operation, as generally innumerable difficulties are associated: the real objects are non-rigid, this is a considerably number of the objects that must be tracked, as in the medical image area, present low rigidity levels; the objects to track can suffer partial or total occlusion along some of the sequence images, or just "disappear" after some images; new objects can appear along the sequence; the objects to track can suffer topology variations, as merge or split operations; etc.
The motion analysis is one of the main tasks in many Computational Vision applications, as it allows the description and quantification of the motion (or the deformation) that an object suffers along the time in an image sequence; that is, it translates the temporal dynamics verified by the object. Generally, this task is the one that appears associated to the object tracking operation. |