Camera self-calibration from multi-view images

Camera calibration is a key technology in computer field, in which self-calibration is to compute camera intrinsic parameters only from a series of images. The Kruppa's equations method not only needs to compute the fundamental matrix which includes all of the geometry relation between images, but also needs to compute the epipoles which variate with the different images. Hartley deduced a simple form of Kruppa's equations in terms of fundamental matrix. The goal of this paper is to convert the equations into the form of cost function according to the Hartley's deduction and calculate the cost function by the sum of fundamental matrixes multiplying the corresponding weighting factors that are related with the proportion of the number of matching features to the image pixels. In this paper, the genetic optimal algorithms are used to achieve the minimum value of the cost function. Our algorithm doesn't require computing the image epipole and avoiding the results instability, and is easy for the calculation. Experimental results show that the proposed algorithm is more effective and accurate, and can become a versatile tool for camera calibration, it will be used in our cameras array calibration.