H. Shariat and K. E. Price. "Motion estimation with more than two frames". IEEE Trans. on Pattern Analysis and Machine Intelligence, 12(5):417--434, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of the horizontal and vertical viewing direction, measured over a larger number of frames. This enables us to exploit temporal smoothing in order to achieve a robust solution for our practical situation. A good overview of the work on egomotion estimation in long image sequences can be found in [10] and [13] We start by defining two coordinate frames: the real world frame and the observer frame. Points in these frames are denoted by (X;Y;Z) and (X ;Y;Z) respectively. The origin of these frames is located at the nodal point of the camera. The z axis of the real world frame points opposite ....

H. Shariat and K. E. Price. Motion estimation with more than two frames. IEEE Trans. on Patt. Anal. and Mach. Intell., Volume 12, Number 5, pages 417--434, MAY 1990.

....sequence is relatively small. The resulting motion and structure estimates from two views alone may be very sensitive to noise and inaccurate. Therefore some of the research in this area concentrates on error analysis and using more than two images from a monocular sequence to improve robustness [9, 48, 49]. Another direction of research is to use a long image sequence; generally there are two approaches: batch and recursive. The batch approach assumes that all the images of a very long sequence are readily available at once so that more data is available, reducing the effects of noise and outliers ....

H. Shariat and K. E. Price, "Motion estimation with more than two frames," IEEE Trans. Pattern Analysis and Machine Vision, 12(5):370--389, 1990.

.... problem has been studied extensively (for example, Ullman[77] Tsai[74] Faugeras Maybank[24] Weng, Ahuja, Huang[79] Consequently, many authors assume that it is solved in their analysis (for example, Broida Chellappa[9] Young Chellappa[82] Burger Bhanu[11] Shariat Price[70]) However, the correspondence problem remains unsolved, in general. Before proceeding with the recovery of 103 CHAPTER 6. TRACKING ACROSS MULTIPLE FRAMES 104 motion and depth estimates, we must develop a tracking mechanism that is applicable to simple region features. It is desirable to obtain ....

....so makes the motion and depth estimation more accurate and robust. It is well known that increasing the distance that the camera moves between frames (i.e. the baseline) in a two frame motion estimation algorithm improves the accuracy of the motion and depth estimates (for example, Shariat Price[70], Heel[30] Michael[56] or Okutomi Kanade[58] However, increasing the baseline also has the effect of increasing the distortion of the image from one frame to the next frame. Therefore, determining correspondence mapping becomes more difficult. Obtaining a correspondence across multiple ....

[Article contains additional citation context not shown here]

H. Shariat and K. Price. Motion estimation with more than two frames. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(5):417--434, May 1990.

....to continuous surfaces of objects in the real world, the segmentation may be considered not only as a tool to estimate discontinuous motion but also as a valuable information in itself. It is worthwhile to note that the motion smoothness constraint can be also extended in the temporal direction [74], 8] 7] or in the direction of motion trajectories [83] An even further extension of this idea is explicit modeling of motion trajectories and estimation of the associated parameters. For example, in [13] motion trajectories are modeled by second order curves (equation (5) and their ....

H. Shariat and K. Price, "Motion estimation with more than two frames," IEEE Trans. Pattern Anal. Machine Intell., vol. 12, pp. 417--434, May 1990.

....areas of intense research in the last decade. Numerous approaches have been reported in the literature. For a review see [1] Problems associated with these approaches, such as high computational complexity, high noise sensitivity, and inefficient use of available information, are enumerated in [2]. We show in this paper that, by working in appropriate coordinate systems and by making use of a priori knowledge about scene geometry and real world constraints on object movements, simple yet robust SFM algorithms can be developed. The motivation of this work comes from the desire to build 3 D ....

H. Shariat and K. E. Price, Motion Estimation with More Than Two Frames, IEEE Trans. Pattern Anal. Mach. Intell., vol.12, 1990, pp.417-434.

....nature of scene components have mostly not yet been investigated. Most works deal with non transitory image sequences, and successful improvements have been achieved in this type of fusion (a partial list: Ayache and Faugeras [5] Matthies and Shafer [9] Tirumalai et al. [12] Shariat and Price [11], Broida and Chellappa [2] Kumar et al. [6] Experiments for scene construction from transitory image sequence only started recently, and we have so far seen two efforts by Cui et al. [3] and Tomasi and Kanade [13] respectively. In Cui et al. [3] some relative accuracy was reported from a ....

H. Shariat, and K. E. Price, Motion estimation with more than two frames, IEEE Trans. Pattern Anal. Machine Intell., vol. 12, no. 5, pp. 417-434, 1990.

.... whereas techniques that infer local shape from image contours presume physical properties of the scene such as smoothness [25] 69] An approach called Structure From Motion (SFM) makes minimal assumptions about the camera and scene and does not require specialized hardware (Shariot and Price [49] and Ullman [67] summarize earlier work in SFM, and Oliensis [39] gives a critique of recent techniques) Ideally, SFM allows arbitrary camera motion and only requires that the scene is static. However, many SFM techniques, such as the Factorization Method and its derivatives, assume a ....

....features. 9 number of camera and feature locations that are consistent with the observed images [66] 64] 20] 29] 14] 74] It is also well understood that these so called minimal data solutions are very sensitive to noise and additional image data must be included for reliability [49], 62] 61] 10] Redundancy over determines the inverse projection problem and therefore projectors no longer precisely intersect due to noise. Projected Error Refinement therefore formulates inverse projection as a optimization problem for determining the positions of the cameras and feature ....

[Article contains additional citation context not shown here]

H. Shariat and K.E. Price, "Motion estimation with more than two frames," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 12, no. 5, 1990, pp. 417-434.

....monitoring, for example, by law enforcement officers. The automatic recognition, tracking, and destruction of moving targets has obvious military interests. The moving object to track can be a car, a truck, a sailing boat, or in military applications a tank, an airplane, or a helicopter etc. [Jain79, Miti88, Shar90, Chen90, Samy84, Nage86b, Leun90, Nich91, Bhan91, Leun91, Seib92, Bhan92b]. Robot vision and vehicle navigation offer applications especially in industry [Aloi90, Bhan92a] Visual quality inspection offers many possibilities to improve the efficiency and the productivity of industrial manufacturing [Huan83] In the field of communications, video signal compression using ....

....for example, Longuet Higgings and Pradzny [Long80] Waxman et al. Waxm88] and Subbarao [Subb87] In addition to Aggarwal s reviews of the 2 D and 3 D motion estimation methods, there are also others articles, with overviews and comparisons, to see recent developments. Shariat and Price [Shar90] have given a survey of motion estimation which uses feature point matches. On the other hand, motion estimation from 3 D line segments can be done with extended Kalman filter, Gauss Newton minimization or singular value decomposition [Zhan91] Heeger and Jepson [Heeg92] categorized 3 D motion ....

Shariat, H., Price, K.E., "Motion Estimation with More Than Two Frames," IEEE Trans. on Pattern Analysis and Machine Vision, vol. 12, no. 5, 1990, pp. 417--434.

....requires that at least 5 independent rigid points are tracked (per target) Mitiche, 1986; Mitiche, 1989) Hence, the general case of 3D rigid motion cannot be recovered from a pair of tracked lights over 2 frames. Whilst many multiple frame motion interpretation algorithms have been developed (Shariat Price, 1990) these impose additional assumptions regarding constancy of (relative) motion which cannot be guaranteed for a system exhibiting independent motion by both an observer and target. Given the intractibilty of solving the general 3D rigid motion case additional constraints are required to allow ....

....an independently moving target. Hence, unlike these traditional approaches, the recovered relative motion parameters cannot be compared in some inertial reference frame. Obviously, since both the sensor and target are moving independently constant relative motion assumptions over multiple frames (Shariat Price, 1990) will not be valid in general, e.g. forward relative translation only. Note that even if relative motion parameters and depths may be solved for, only the direction of relative translation may be recovered, due to the depth scale and depth velocity ambiguities (Matthews, 1992) Since no odometric ....

Shariat, H., & Price, K.E. 1990. Motion estimation with more than two frames.

....to continuous surfaces of objects in the real world, the segmentation may be considered not only as a tool to estimate discontinuous motion but also as a valuable information in itself. It is worthwhile to note that this smoothness constraint can be also extended in the temporal direction [73, 8, 7] or in the direction of motion trajectories [82] An even further extension of this idea is explicit modeling of motion trajectories and estimation of the associated parameters. For example, in [13] motion trajectories were modeled by second order curves (equation (5) and their parameters ....

H. Shariat and K. Price, "Motion estimation with more than two frames," IEEE Trans. Pattern Anal. Machine Intell., vol. 12, pp. 417--434, May 1990.

....for example by law enforcement officers. The automatic recognition, tracking and destruction of moving targets has obvious military interests. The moving object to track can be a car, a truck, a sailing boat, or in military applications a tank, an airplane, or a helicopter etc. 53] 85] [125], 27] 119] 94] 71] 95] 20] 72] 121] 22] Robot vision and vehicle navigation offer applications especially in industry [12] 21] 123] For example visual quality inspection offers many possibilities to improve the efficiency and the productivity of industrial manufacturing [49] In the ....

....algorithm which minimizes the proximal uniformity cost function and establishes correspondence over n frames. The algorithm was compared to corresponding methods. They had also improved their method extending it to the use of optical flow to get the initial correspondence. 27 Shariat and Price [125] gave a survey of motion estimation using feature point matches. The survey of past results shows different variants having their own characteristic limitations like the number of points, the type of equations (linear, nonlinear, etc. They suggested three cases of their own to detect motion ....

Shariat, H., Price, K.E., "Motion Estimation with More Than Two Frames", IEEE Transactions on Pattern Analysis and Machine Vision, vol. 12, no.5, 1990, pp. 417-434

No context found.

H. Shariat and K. E. Price. "Motion estimation with more than two frames". IEEE Trans. on Pattern Analysis and Machine Intelligence, 12(5):417--434, 1990.

No context found.

H. Shariat and K. Price. Motion estimation with more than two frames. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(5):417--434, May 1990.

No context found.

H. Shariat and K.E. Price. Motion estimation with more than two frames. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(5), May 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC