### Table 1: Analysis of the methods of blood vessels visualization according to the physicians requirements. with the stationary tissue and with zooming of the areas of interest. The volume rendering techniques are in general slower than the MIP ones and than the surface rendering approach, but they give 3D images than do not change the vascular topology and that emphasize the perception of the vessels on the basis of the classi cation results. With symbolic models, as the reconstruction of the surface is enabled while keeping the voxel data, the volume rendering approach allows to obtain better quality integrated images. Thus their optimization is a challenging alternative to the other methods. In [Pui98] a symbolic model of the vascular map allowing the representation of its topology, its surface and its volume information is presented. Next, this model is brie y described and several strategies for its visualization are analyzed.

1998

"... In PAGE 6: ... In order to reference the blood vessels in relation to the stationary tissue [HTL+89] proposes to perform a mixed ray-casting that detects the maximum value along the ray and then integrates the intensity from this point, along the ray of vision. In Table1 the methods analyzed are classi ed according to the requirements exposed above. Synthesizing, on one hand the MIP technique does not keep the topology of the vessels, nor the tridimensionality of the data, and it can produce artifacts.... ..."

Cited by 1

### Table 2. Quantitative comparison between the computed 3D spine curve and axial vertebral rotation and ground truth data. The results are shown separately for T1-andT2-weighted images. Overall mean values and the corresponding standard deviations are shown in the bottom rows. Spine segments Whole spines

2007

### Table 2: Number of classified segments with rotation of the image

### Table 2: Twenty Statistical Measurements on Segmented Image

1993

"... In PAGE 4: ...indow quot; in which it is centered. Later we will apply GP to the classification of these features. In the conventional MTAP algorithm, however, they are passed to a second processing stage. F00 Size Filter Value ( quot;Blob Contrast quot;) F01 Global Image Intensity Mean F02 Global Image Intensity Standard Deviation F03 quot;Large Window quot; Intensity Mean F04 quot;Large Window quot; Intensity Standard Deviation F05 quot;Small Window quot; Intensity Mean F06 quot;Small Window quot; Intensity Standard Deviation Table2 : Seven primitive features are based on a hierarchy of image window intensities. 1 3.... In PAGE 5: ...he terminal set T = {F00...F19, RANFLOAT} for the first experiment consists of the 20 segmented quot;statistical quot; features shown in Table2 as well a real random variable RANFLOAT, which is resampled to produce a constant value each time it is selected as a terminal node. The resulting tree takes a bag of floating-point feature values and constants as input, and combines them through linear and nonlinear operations to produce a numeric result at the root of the tree.... ..."

Cited by 81

### Table 2: Evaluation of Region Segmentation on Standard Clips

"... In PAGE 5: ... A still image from each of these is shown in Figure 2. A value for Temporal Coherence (Tcoh) for each sequence was computed based on segmenting each frame of each of the four sequences independtly of neighbouring frames and this gives the values for the Independent frames in Table2 . A coherent regional segmentation for the sequences was estimated twice, firstly based on an estimation of frame coherency against frames in the original sequence only ( first iteration ) and secondly based on an estimation of frame coherency against frames newly gener- ated in lieu of the original frames ( second iteration ).... In PAGE 7: ... For this reason the absolute values for Tcoh for one video sequence cannot be compared to Tcoh values from different sequences and the appropriate application of the mea- sure is to compare the temporal coherence of different segmentation algorithms on the same sequence. Thus the values for Tcoh in Table2 should be compared for segmentations within each sequence and not across different sequences. Acknowledgments This work is partly supported by Science Foundation Ireland under grant 03/IN.... ..."

### Table 1: Comparison Between Segmentation Techniques Segmentation Technique Information used in Segmentation Extends to N-D Method of Segmentation

"... In PAGE 6: ... Boykov and Jolly (2001a) use gray-scale histograms in order to model the intensities of the pixels in the image, however the Grab-Cut technique uses Gaussian mixture models in order to model the colour statistics of the image. Table1 shows a comparison between the various segmentation techniques with regard to the information they use to perform the segmentation, if the method can be extended from 2-D to N-D data, and the method that is employed in order to perform the segmentation. Segmentation techniques are used primarily for 2-D images, but some segmentation techniques can be extended to 3-D data (or volumes) and even N-D data.... ..."

### Table 1: The two displacements used to illustrate the recovery algorithms. Note: The motion is speci ed in the rst camera coordinate system (X and Y axis aligned with the image coordinate system). The rst displacement is dominated by a translation perpendicular to the image plane, and the second displacement by a translation parallel to the image plane. To simulate a general motion, we have also added a rotational component. The axis of rotation is the direction of the rotation vector, and its angle is the norm of that vector. rotation translation correct value of epipole

1998

"... In PAGE 13: ...Two typical displacements The experimental part of this chapter is illustrated with two examples of typical displacements whose values are shown as Table1 . The rst displacement yields epipoles within the image, whereas the second displacement yields epipoles that lie far outside.... In PAGE 34: ...Table1 0: Statistic: Average relative distances obtained with the cross-ratio method. C: normalization by the covariance, P: applying the permutations.... In PAGE 35: ...Table1 1: Summary of methods tried manipulated quantity obtained with operations complexity unknowns triplets of epipoles 7 points clustering closed form algebraic 2 plane cubics 6 points approximative intersection deterministic iterative 2 cross-ratios 4 points non-linear minimization non-deterministic iterative 4 seems to be generally no: used alone, the Fundamental matrix method generally gives a better precision. We have delimited the applicability of the direct methods.... ..."

Cited by 5

### Table 2 presents experimental results comparing real and measured va- riations with imbeded spheres, imbeded iso-surfaces and segmentation. The segmentation underestimates the volume variation because the synthetic lesion is fussy: the intensity of its boundary is very similar to the one of the under- lying image. In that case, the deformation eld method gives slightly better results than segmentation. Figure 13 presents an example of measured pro le.

"... In PAGE 25: ... Table2 : Synthetic experiments for central deformation. 6.... ..."

### Table 1. Comparison of geometry (width and amplitude distributions) on vascular segmentations using MSF and RTM methods and ground-truth images from STARE database using KLD. The more negative the number the worse the comparison. RTM is superior in all but one instance.

### Table 1. Optimization techniques used in image segmentation and multiview 3D re- construction.

2007

"... In PAGE 3: ... Both techniques were inspired by the original works of [13], [24]. Table1 provides a number of representative works on local opti- mization, discrete and continuous global optimization in the context of image segmentation and multiview reconstruction, respectively. The paper is laid out as follows.... ..."

Cited by 1