Predicting a dense depth map from synchronized LiDAR scans and RGB images using compact deep neural networks presents significant challenge. While most state-of-the-art models enhance prediction accuracy by increasing the number of parameters, leading to substantial memory consumption, completion tasks in areas such as autonomous driving primarily utilize edge devices powered embedded GPUs. In ...

We propose a low-complexity algorithm for stereoscopic video applications that generates a high-quality 3D image depth map from a single 2D image. Based on their characteristics, 2D images are classified into one of three categories before being processed by the proposed lowcomplexity algorithm to generate corresponding depth maps. We also extend the 3D depth algorithm to construct a parallel 3...

We propose a framework for extracting structure from stereo which represents the scene as a collection of approximately planar layers. Each layer consists of an explicit 3D plane equation, a colored image with per-pixel opacity (a sprite), and a per-pixel depth offset relative to the plane. Initial estimates of the layers are recovered using techniques taken from parametric motion estimation. T...

Complex Gabor filters with phases in quadrature are often used to model evenand odd-symmetric simple cells in the primary visual cortex. In stereo vision, the phase difference between the responses of the left and right views can be used to construct a disparity or depth map. Various constraints can be applied in order to construct smooth maps, but this leads to very imprecise depth transitions...

We generalize the QSQR evaluation method to give a setoriented depth-first evaluation method for Horn knowledge bases. The resulting procedure closely simulates SLD-resolution (to take advantages of the goal-directed approach) and highly exploits set-at-a-time tabling. Our generalized QSQR evaluation procedure is sound, complete, and tight. It does not use adornments and annotations. To deal wi...

This paper presents an e cient technique for unsupervised content-based segmentation in stereoscopic video sequences by appropriately combined di erent content descriptors in a hierarchical framework. Three main modules are involved in the proposed scheme; extraction of reliable depth information, image partition into color and depth regions and a constrained fusion algorithm of color segments ...

3DTV gets more and more popular. However, with the growing number of stereoscopic display devices, the 3D content for display is obviously insufficient. Current methods for generating single camera stereoscopic videos are based on creating a depth map. In this paper, an efficient image depth extraction method which utilizes the support vector machine(SVM) process is proposed. The label is estab...

Given a pair of stereo images, the spatial coordinates of a scene point can be derived from its projections onto the two considered image planes. Finding the correspondences between such projections however remains the main difficulty of the depth estimation problem: the matching of points across homogeneous regions is ambiguous and occluded points cannot be matched as their projections do not ...

This paper surveys the planning and reasoning research being carried out in the Robot Vision Lab at Purdue. In parlicular, we will describe the workings of a new planning system called SPAR, which uses a constraint posting approach for simultaneously fulfilling the operational, geometric and uncertainly reduction goals, and the PSEIKI system for evidential reasoning in a rangled hierarchy. We w...

This paper describes a method of extracting depth information about a scene from a single, static viewpoint. The. approach uses aperture variation to obtain a sequence of images differing only in depth of field, thus avoiding the 'correspondence problem associated with stereo and optical flow techniques. These images contain the necessary coded depth information. By using one image (that with t...