Active contours: Generalization of the snake mode

To read the full-text of this research, you can request a copy directly from the authors.


— Segmentation is one of the fundamental issues in the field of image processing and computer vision. Various approaches include differentiating an object in the image as a final goal or for further processing (medical diagnosis, surveillance, 3-D reconstruction and more). Snakes, a model proposed by Kass, Witkin, and Terzopoulos in 1987, provides an efficient method for segmenting an object through the minimization of its energy. The advantage of snakes is in its ability to use high-level data given by the algorithm operator, as opposed to other methods such as the Laplace technique. The snakes model inherently imposes strong constraints on a given image in order to successfully segment an object. In this paper, the use of adjustment methods is described, which allow us to generalize the snake model to a wider range of applications. Through the use of pre-processing techniques, the model's constraints were softened. The main theoretical model and its use in facing a real life image is presented.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

ResearchGate has not been able to resolve any citations for this publication.
In most display applications, the final step is the viewing by human observers, so it is only logical to use current quantitative knowledge of the visual system in the design of imaging and display algorithms. This talk will cover key human visual system attributes, their modeling as separate components, and how they have been used in our labs to design new display algorithms and fine-tune existing approaches. Some of the resulting algorithms described will include: motion-adaptive backlight flashing, skin-cognizant color mapping, two-spatial channel architecture for color mapping and decontouring, bit-depth extension, subpixel subsampling, and wavelet compression. The talk will also cover the new challenges that arise in very large (>100″) and bright (>500 cd/m∧2) displays, in terms of peripheral sensitivity and improved smooth pursuit eye movement capability.
— This paper presents an approach for the compression of mixed-image data (images which include text) using a segmentation-based image representation. The mixed image to be compressed is approximated as a set of uniform regions encoded as a set of region boundaries (contours) and descriptions of the region interiors. In order to improve the visual quality of textual portions in the image approximation, coarsely quantized approximation error (corresponding only to the textual portions) is also encoded. The segmentation method used to form this representation combines a Gaussian texture model and a Gibbs-Markov contour model in order to find regions with boundaries which correspond closely to the objects in the image. A mixed image, in general, is difficult for conventional transform coding methods because these methods typically discard high-frequency components or quantize them coarsely. Structures such as edges in the image, however, are dependent upon these high-frequency components. Therefore, the visual quality of structures such as text (which consists mostly of high-contrast edges) in the image is severely degraded by conventional transform coding methods. Due to its ability to capture the semantically important structure in the image, segmentation-based coding is a promising approach for these images. The effectiveness of the proposed approach for the compression of mixed images is demonstrated through experimental results.
— While the need for image-sequence compression has long been understood, recent applications have shown that existing methods are not always sufficient. A case in point is the class of applications referred to as “very low bit rate.” These applications require the transmission of sequences over very narrow channels, e.g., 9600 bits/s. The compression ratios required are substantially beyond those that can be accomplished (with reasonable image quality) using standard methods. The Gabor transform is a local frequency decomposition which has been proposed as a model of the early stages of human vision. It has found widespread use for a number of tasks in signal and image processing, including image compression. In this paper, we examine Gabor transform-based coding for image-sequence transmission at rates near 10 kbits/s.
A snake is an energy-minimizing spline guided by external constraint forces and influenced by image forces that pull it toward features such as lines and edges. Snakes are active contour models: they lock onto nearby edges, localizing them accurately. Scale-space continuation can be used to enlarge the capture region surrounding a feature. Snakes provide a unified account of a number of visual problems, including detection of edges, lines, and subjective contours; motion tracking; and stereo matching. We have used snakes successfully for interactive interpretation, in which user-imposed constraint forces guide the snake near features of interest.
Optimal code for image processing applications
  • C Sagiv