MDL-based spatiotemporal segmentation from motion in a long image sequence

Gu, Xianfeng; Shirai,; Asada,

doi:10.1109/cvpr.1994.323865

Cited by 9 publications

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The algorithm combines aspects of snakes and region growing, and is guaranteed to converge to a local minimum. In related work, [23] presented a method for spatiotemporal segmentation of long sequences of images based on the MDL principle and simultaneously obtained optimal spatial segmentation and motion estimation without extracting the optic-flow field.…”

Section: Minimum Description Length Principlementioning

confidence: 99%

Corrections to "Deformable shape detection and description via model-based region grouping"

Sclaroff

Liu²

2001

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

ÐA method for deformable shape detection and recognition is described. Deformable shape templates are used to partition the image into a globally consistent interpretation, determined in part by the minimum description length principle. Statistical shape models enforce the prior probabilities on global, parametric deformations for each object class. Once trained, the system autonomously segments deformed shapes from the background, while not merging them with adjacent objects or shadows. The formulation can be used to group image regions obtained via any region segmentation algorithm, e.g., texture, color, or motion. The recovered shape models can be used directly in object recognition. Experiments with color imagery are reported.

show abstract

Section: Minimum Description Length Principlementioning

confidence: 99%

Corrections to "Deformable shape detection and description via model-based region grouping"

Sclaroff

Liu²

2001

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

show abstract

“…For automatic extraction of partial contours, various methods of edge constitution have also been reported. Some methods approximate and constitute the edge pixels using curves or ellipses [12], some methods minimize the spatial position of edge pixels using curves [13], and some methods perform spatiotemporal segmentation by minimizing the motion of edge pixels in a long image sequence [14]. In these methods, tracking of contours has not been considered, so that correspondence of contours between images becomes a problem.…”

Section: Introductionmentioning

confidence: 99%

Automatic extraction and tracking of complex contours

et al. 1999

View full text Add to dashboard Cite

In this paper, we propose an algorithm to extract and track complex contours with occlusions such as the contours of the human body. In this algorithm, complex contours are assumed to be composed of contour segments which have smooth variation of the intensity gradient normal to the contour direction. Moreover, the contour segments are modeled using B splines, and contours in the image are extracted and tracked by the contour segments. In the first frame, contour segments are extracted using only the edge pixels. In the tracking of contour segments of the next frame, we use Snakes. For frames after the second frame, the contour segments are redetected using the tracking result and edge pixels. After that, extraction and tracking are repeatedly applied to each frame. Moreover, a method of classification of contour segments which is based on movement is described and the possibility of human action description is investigated. Finally, experimental and classification results obtained by means of this algorithm on synthesized and real images are given. ©1999 Scripta Technica, Syst Comp Jpn, 30(8): 40–50, 1999

show abstract

“…They find the motion parameters of a human body part by calculating the optical flow of several points within this part and compare them to the movements of a model of the part. Gu et al [48] instead track edge features, defined by their length and contrast, in consecutive images using the optical flow constraint. In the work by Bregler [17] each pixel is represented by its optical flow which is grouped into blobs having coherent motion and represented by a mixture of multivariate Gaussians.…”

Section: Temporal Datamentioning

confidence: 99%