The quadrifocal variety (Q332657)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The quadrifocal variety |
scientific article |
Statements
The quadrifocal variety (English)
0 references
8 November 2016
0 references
Perhaps the most important problem in computer vision is that of scene reconstruction of a real \(3\)-dimensional world scene from a given set of \(2\)-dimensional images or projections of it. Multiple view geometry, from an algebraic geometry perspective, is one of several approaches to this problem. The paper under review gives a quick survey of multi-view geometry from an algebraic geometry perspective: Section 3 reviews multi-focal tensors via equivariant projections of a Grassmannian, emphasizing connections with previous work of [\textit{R. Hartley} and \textit{F. Schaffalitzky}, 363--375, in Computer Vision, ECCV 2004. Berlin: Springer (2004; Zbl 1098.68775)] and [\textit{O. Faugeras} and \textit{B. Mourrain}, ``On the geometry and algebra of the point and line correspondences between \(n\) images'', in Proceedings of the Fifth International Conference on Computer Vision. Cambridge, Ma: IEEE. 951--956 (1995)]. Section 4, on arbitrary dimensional cameras, highlights a relation between the multi-focal variety and the variety of principal minors of square matrices, now with connections with previous work of [\textit{S. Lin} and \textit{B. Sturmfels}, J. Algebra 322, No. 11, 4121--4131 (2009; Zbl 1198.14049)]. The main results of the paper are in Sections 5 and 6, where using methods from representation theory and symbolic computational algebra and algebraic geometry, in particular, using that the quadrifocal variety is invariant under the action of the group \(G=S_4 \ltimes (\text{GL}_3)^4\), which means that its corresponding ideal can be described as a \(G\)-module, the author determines generators for the ideal of the quadrifocal variety up to degree \(8\) with partial results in degree \(9\). The corresponding results and computations of Section 6 are too involved to be included in this review.
0 references
computer vision
0 references
quadrifocal tensor
0 references
multi-view geometry
0 references
0 references