Author: eric2i

  • Image Abstraction using Layer-wise Image Vectorization

    LIVE is a solid research work released in June 2022 (accepted by CVPR 2022 oral) regarding image vectorization, the reverse process of image rasterization, which is also an important problem in Computer Graphics.

    At the end of my Digital Image Processing course, I  used this algorithm to generate a vectorized version of my selfie! It is also possible to create a short video to demonstrate the process of this algorithm.

    A vectorized version of my selfie with path=64.

    Besides simple context, LIVE can also be adapted to complex scenic images, such as the following photo I took at the Beijing University of Chemical Technology (BUCT).

    A vectorized version of one BUCT campus image with path=256.
  • Magnetic Field Intensity Around a Uniformly Carrying Cylindrical Surface

    In general, by utilizing Ampère’s Circuital Law, it is straightforward to derive the formula for the magnetic induction intensity around a uniformly current-carrying cylindrical surface, except on the surface of the cylinder itself:

    $$B(r)=\begin{cases}
    0 & if 0 < r < R \\
    \frac{\mu_0 I}{2\pi r} & if r > R \\
    \end{cases}$$

    For the case where \(r=R\), Ampère’s Circuital Law cannot compute the value.

    Instead, we use the Biot-Savart Law, treating the infinitely long cylindrical surface as an infinite straight current. The surrounding field strength formula is then:

    $$B(r)=\frac{\mu_0 I}{2\pi d} $$

    where \(i=\frac{ds}{2\pi R}I\).

    It is intuitive to place the research point on the cylindrical surface at the coordinate origin and use polar coordinates to rewrite the equation. Given that:

    $$ds = \sqrt{\rho^2 + \rho^{‘2}} = 2R\theta$$

    $$d=\rho$$

    Thus, the differential magnetic field \(dB\) is given by:

    $$dB = \frac{\mu_0 I}{4\pi^2 R}d\theta$$

    Integrating from \(-\pi/2\) to \(\pi/2\), we find:

    $$B = \int_{-\pi/2}^{\pi/2}\frac{\mu_0 I}{4\pi^2 R}d\theta = \frac{\mu_0 I}{4\pi R}$$