The varieties of bifocal Grassmann tensors (Q6093331): Difference between revisions

The paper under review is motivated by the subject of ``multiple view geometry'' (sometimes also called ``multi-view geometry''), which is a currently very active branch of computer vision. Scientists working in multiple view geometry have been interested in certain algebraic varieties that naturally come up in their study; the paper under review provides a thorough mathematical analysis of certain of these varieties. Specifically, multiple view geometry deals with the reconstruction of an object in a projective space \({\mathbb P}^k\), given \(r\) projections of the object to lower-dimensional projective spaces \({\mathbb P}^{h_1},\ldots,{\mathbb P}^{h_r}\) (in the dictionary linking computer vision and mathematics, a projection from \({\mathbb P}^k\) to \({\mathbb P}^{h_r}\) is called a ``camera'', underlining the connection with photography and video). The algebraic varieties arising in this setting are the varieties parametrizing Grassmann tensors (also known as multifocal tensors), which have been introduced in [\textit{R. I. Hartley} and \textit{F. Schaffalitzky}, Int. J. Comput. Vis. 83, No. 3, 274--293 (2009; Zbl 1477.68473)]. The paper under review is about the special case \(r=2\) (i.e. 2 cameras), and the varieties arising here are known as the varieties of bifocal Grassmann tensors. The first main result (Theorem 4.1 in the paper under review) is that the variety of bifocal Grassmann tensors is birational to a certain homogeneous space. The second main result (Theorem 4.5 in the paper) gives a more precise description of the structure: the variety of bifocal Grassmann tensors admits a dominant rational map (which is birationally equivariant) to a certain Grassmannian. The paper is clear and well-written, with several concrete and explicit examples that guide the reader. A reader unfamiliar with computer vision might further benefit from perusing the classical reference [\textit{R. Hartley} and \textit{A. Zisserman}, Multiple view geometry in computer vision. With foreword by Olivier Faugeras. 2nd edition. Cambridge: Cambridge University Press (2004; Zbl 1072.68104)], as well as earlier work by the authors of the paper under review [Ann. Mat. Pura Appl. (4) 196, No. 2, 539--553 (2017; Zbl 1367.15046); SIAM J. Matrix Anal. Appl. 41, No. 2, 591--604 (2020; Zbl 1505.15022))].