Bicubic interpolation

About: Bicubic interpolation is a research topic. Over the lifetime, 3348 publications have been published within this topic receiving 73126 citations.

Accuracy of Multiply-Upstream, Semi-Lagrangian Advective Schemes

Abstract: The stability and accuracy of the multiply-upstream, semi-Lagrangian method of integrating the advective equation in two dimensions is examined for four different interpolation schemes; namely, bilinear, biquadratic, bicubic and biquartic. All are shown to be consistent and unconditionally stable for constant advecting velocity. Their respective amplitude and phase errors are discussed. They are then used to integrate the test problem of a cone being advected about the plane at constant angular velocity. The merits of the schemes relative to each other and relative to a well tried Eulerian scheme am examined with particular regard to accuracy and computation time.

Fourier reconstruction in optoacoustic imaging using truncated regularized inverse k-space interpolation

Abstract: A novel Fourier transform based reconstruction algorithm for solving the inverse problem in optoacoustic imaging is presented, which improves reconstruction efficiency and image quality. Fourier algorithms make use of an interpolation law when signal Fourier components are mapped to source Fourier components. To overcome inadequacies affiliated with interpolation methods such as nearest neighbour, linear, cubic or spline interpolation, together with signal data zero padding, we present a regularized interpolation method based on a forward model explicitly formulated for the compactly supported signal data. Simulations performed on a digital tissue phantom reveal the potential of this novel reconstruction method, which results in images of enhanced quality but without the need of using time-consuming zero-padding.

Cycle-Dehaze: Enhanced CycleGAN for Single Image Dehazing.

Abstract: In this paper, we present an end-to-end network, called Cycle-Dehaze, for single image dehazing problem, which does not require pairs of hazy and corresponding ground truth images for training. That is, we train the network by feeding clean and hazy images in an unpaired manner. Moreover, the proposed approach does not rely on estimation of the atmospheric scattering model parameters. Our method enhances CycleGAN formulation by combining cycle-consistency and perceptual losses in order to improve the quality of textural information recovery and generate visually better haze-free images. Typically, deep learning models for dehazing take low resolution images as input and produce low resolution outputs. However, in the NTIRE 2018 challenge on single image dehazing, high resolution images were provided. Therefore, we apply bicubic downscaling. After obtaining low-resolution outputs from the network, we utilize the Laplacian pyramid to upscale the output images to the original resolution. We conduct experiments on NYU-Depth, I-HAZE, and O-HAZE datasets. Extensive experiments demonstrate that the proposed approach improves CycleGAN method both quantitatively and qualitatively.

Smooth surface approximation to serial cross-sections

Abstract: The reconstruction of the surface model of an object from 2D cross-sections plays an important role in many applications. In this paper, we present a method for surface approximation to a given set of 2D contours. The resulting surface is represented by a bicubic closed B-spline surface with C2 continuity. The method performs the skinning of intermediate contour curves represented by cubic B-spline curves on a common knot vector, each of which is fitted to its contour points within a given accuracy. In order to acquire more compact representation for the surface, the method includes an algorithm for reducing the number of knots in the common knot vector. The proposed method provides a smooth and accurate surface model, yet realizes efficient data reduction. Some experimental results are given using synthetic and MRI data.