Radon transform, and more specifically the multiscale Discrete Radon Transform, is a valuable tool to find straight structures on images.
But the straight lines that DRT finds are those traversing completely the image, and hence DRT is unable to detect non-straight structures.
That limitation is normally adressed by evaluating Radon transform on non-overlapped subsquares of the image. In this work we will show that the initial partial transform stages of multiscale DRT contain enough information to detect non-straight edges.
By stopping on initial stages of an already linearithmic transform that can be executed on integer arithmethics, the proposed method is found to be much faster than its alternatives.
We will show its ability to process images at video acquisition rate on mobile phones.
Points of view generation allows to create virtual views between two or more cameras observing a scene. This field receives attention from multimedia markets, because sufficiently realistic points of view generation should allow to navigate freely between otherwise fixed points of observation. The new views must be interpolated between sampled data, aided by geometrical information relating real cameras poses, objects in the scene and desired point of view. Normally there are several steps involved, globally known as Structure from Motion (SfM) method. Our study focuses on the last stage; image interpolation based on the disparities between known cameras. In this paper, a new method is proposed that uses depth maps generated by a single camera, named SEBI, allowing a more efficient filling in presence of occlusions. Occlusions are considered during interpolation, creating an occlusion-map and an uncertainty-map using the depth information that SEBI cameras provide.
In this work we have presented a brief insight into the capabilities of multilayer displays as to selectively display information in relation to the observers. We labeled the views of a light-field as blocked and non-blocked, and then a predefined text was assigned accordingly, modifying it to achieve a privacy criterion in the blocked case. Two ways to define the private views were presented. An evaluation of the output for both techniques was carried over in simulation, in both the spatial and frequency domain. Results showed that privacy was achievable and that each technique had an optimal operation point when taking into account the time-multiplexing capabilities of the multilayer display. Also, a trade-off between the quality of the blocked and non-blocked views was found.
Tensor display is an option in glasses-free three-dimensional (3-D) display technology. An initial solution has to be set to decompose the light-field information to be represented by the system. We have analyzed the impact of the initial guess on the multiplicative update rules in terms of peak signal-to-noise ratio, and proposed a method based on depth map estimation from an input light field. Results from simulations were obtained and compared with previous literature. In our sample, the initial values used have a large influence on results and convergence to a local minimum. The quality of the output stabilizes after a certain number of iterations, suggesting that a limit on such numbers should be imposed. We show that the proposed methods outperform the pre-existing ones.