Orthographic View Images Research Articles

Two- and three-dimensional deep human detection by generating orthographic top view image from dense point cloud

Human detection still suffers from occlusion, complex backgrounds, and scale-variant problems. Projecting three-dimensional (3D) points onto the ground to generate an orthographic top view (OTV) image for detection can effectively alleviate these problems. However, depth sensors may be placed arbitrarily, making it difficult to create OTV images by the dense point cloud converted from a depth image. We focus on the generation of OTV images and human detection via the constructed OTV image. First, we propose a ground plane extraction method that is well suitable for various camera positions and orientations in complex scenes. Next, points are converted to a uniform coordinate system by ground parameters and encoded to generate a three-channel OTV image. Then, the mainstream two-dimensional (2D) network is employed to detect the human directly on OTV images and further obtain the 3D bounding box by computing the mapping from the OTV image. Besides, we propose a semiautomated annotation method to solve the problem of few OTV image annotations. The proposed method is evaluated on the EPFL dataset, including two subsets, and achieves state-of-the-art performance compared with the existing approaches. Moreover, our 2D and 3D human detection method can run more than 26FPS on the CPU.

Advanced Three-Dimensional Visualization System for an Integral Imaging Microscope Using a Fully Convolutional Depth Estimation Network

In this paper, we propose an advanced three-dimensional visualization method for an integral imaging microscope system to simultaneously improve the resolution and quality of the reconstructed image. The main advance of the proposed method is that it generates a high-quality three-dimensional model without limitation of resolution by combining the high-resolution two-dimensional color image with depth data obtained through a fully convolutional neural network. First, the high-resolution two-dimensional image and an elemental image array for a specimen are captured, and the orthographic-view image is reconstructed from the elemental image array. Then, via a convolutional neural network-based depth estimation after the brightness of input images are uniformed, a more accurate and improved depth image is generated; and the noise of result depth image is filtered. Subsequently, the estimated depth data is combined with the high-resolution two-dimensional image and transformed into a high-quality three-dimensional model. In the experiment, it was confirmed that the displayed high-quality three-dimensional model could be visualized very similarly to the original image.

Three-area-array coherent-dispersion stereo-imaging spectrometer.

A coherent-dispersion stereo-imaging spectrometer is presented, which combines three-view stereo imaging, interferometric spectroscopy and dispersive spectroscopy. Three area-array detectors record three spectral images of each scene unit from three views. For each of the three views, each scene unit is imaged on a given column of one area-array detector, and different wavelengths are dispersed across different rows of that column. For each scene unit, multiple interferograms are simultaneously generated at each view, each interferogram covering a separate wavelength range and located in a separate pixel. The orthographic view image is used to create a two-dimensional orthophoto image. The front view and back view images are used to reconstruct the three-dimensional stereoscopic image. Preliminary theoretical calculations are given. The instrument is a unique concept to obtain three-dimensional spatial information and one-dimensional spectral information while achieving high spectral resolution measurement of an ultraviolet-visible broadband spectral range (e.g., 0.05 nm at 450 nm together with 0.1 nm at 700 nm). It will be suitable for ultraviolet-visible hyperspectral remote sensing.

Mobile integral imaging display using three-dimensional scanning

A three-dimensional mobile display based on computergenerated integral imaging technique for the real-world object using the three-dimensional scanning is proposed. The real threedimensional information of the object is acquired by the threedimensional scanning; then, the specification of the lens array is inputted by the user, and the acquired and inputted information are organized for the future processing. According to the acquired information, the virtual three-dimensional model for the object is generated, and the virtual space including a virtual model, a lens array, and imaging plane is configured. After the elemental image array is generated from the virtual model, the orthographic-view image and depth-slices are reconstructed and displayed on the mobile display according to the user interaction. The data organizing and image rendering processes are proceeded through the cloud computing due to the limited performance of the mobile device. The theory is approved experimentally, and the experimental results confirm that the proposed method can be an efficient way to display the three-dimensional visualization of the real-world object on the mobile display.

Computational integral imaging reconstruction of perspective and orthographic view images by common patches analysis.

A novel method to computationally reconstruct perspective and orthographic view images with full resolution of a recording device from a single integral photograph is proposed. Firstly, a group of image slices that contain full yet redundant information to reconstruct the view image are generated, and the object surface is divided into pieces by the points that correspond to the centers of image slices. Secondly, the image slices that contribute to the pieces are extracted and redundant information embedded in them are figured out by common patches analysis. Finally, the view image is reconstructed by excluding the redundant information and resampling with maximum sampling rate. Each piece of the object surface is represented with 9 patches at most from 4 adjacent elemental images, and view images with high quality are reconstructed. Both simulations and experiments verify the validity of the method.

P‐80: Multiple Orthographic Image Interleaving for Generating Tilted Elemental Image Array with an Arbitrary Angle Directly

In this paper, we propose a computer‐generated integral imaging (CGII) method called multiple orthographic image interleaving (MOII) that used directly for generating tilted elemental image array (EIA) with an arbitrary angle. An orthographic view image array is obtained by rendering a three‐dimensional (3D) scene at certain positions inside a circle. After that, a pixel mapping algorithm is used to interleave the rendering results to form the tilted EIA with an arbitrary angle. The feasibility and efficiency of the proposed method are verified and analyzed through experiments. The experiment results show that the reconstructed 3D images have clear edge and have no distortion.

Multiple elemental image mapping for resolution-enhanced orthographic view image generation based on integral imaging

In this paper, we propose an orthographic view image generation method based on integral imaging called multiple elemental image mapping. The proposed method is based on matching regions in multiple elemental images, which are generated with different micro-lens from the same area of a 3D object. By mapping all the matching regions to the corresponding orthographic view plane, we can enhance the resolution of the generated orthographic view image. Simulation experiments are carried out, and the results show the resolution of orthographic view image is improved obviously.

Real-time interactive display for integral imaging microscopy.

A real-time interactive orthographic-view image display of integral imaging (II) microscopy that includes the generation of intermediate-view elemental images (IVEIs) for resolution enhancement is proposed. Unlike the conventional II microscopes, parallel processing through a graphics processing unit is required for real-time display that generates the IVEIs and interactive orthographic-view images in high speed, according to the user interactive input. The real-time directional-view display for the specimen for which 3D information is acquired through II microscopy is successfully demonstrated by using resolution-enhanced elemental image arrays. A user interactive feature is also satisfied in the proposed real-time interactive display for II microscopy.

Parameter analysis of integral Fourier hologram and its resolution enhancement

We present a parameter analysis of the integral Fourier hologram that is generated from multiple orthographic view images of a three-dimensional object. The maximum view angle, the lens array pitch, and the projection angle step are analyzed to reveal their effects on the maximum size of the reconstructed object and its maximum spatial frequency. With these analyses, we propose a lens array shift method to enhance the resolution of the reconstructed object from the Fourier hologram. The principles are verified by computational and optical experiment.

Resolution-enhanced integral imaging microscopy that uses lens array shifting

A resolution-enhanced integral imaging microscope that uses lens array shifting is proposed in this study. The lens shift method maintains the same field of view of the reconstructed orthographic view images with increased spatial density. In this study, multiple sets of the elemental images were captured with horizontal and vertical shifts of the micro lens array and combined together to form a single set of the elemental images. From the combined elemental images, orthographic view images and depth slice images of the microscopic specimen were generated with enhanced resolution.

Towards direct transformation of orthographic-view drawings into a prototype

Since its early days, geometrical reconstruction (GR) has been developed as a stand-alone process for reconstructing a 3D CAD model from a 2D engineering drawing. Addressed in this paper is an attempt to extend GR application to interface it with the rapid prototyping process (RP) in order to construct a prototype directly from an engineering drawing. A layer-based geometric reconstruction (LBGR) concept has been developed to couple the two applications. LBGR locally translates information on the 2D drawing into a stack of contours section-by-section in a top-down direction, in which the top and bottom contours of each section are identified before its intermediate contours are generated. It can be applied to parts that consist of regular surfaces whose edges are clearly presented on the orthographic views (e.g. plane surfaces, conical surfaces, ruled surfaces with cross-sectional contours parallel with the xy plane). This concept has been implemented in LabVIEW to demonstrate this direct interface. Inputs and outputs are orthographic view images of a drawing and the vector files of contours, respectively. The direct transformation of mechanical part drawings into physical prototypes is also presented in this paper.

Fresnel and Fourier hologram generation using orthographic projection images

A novel technique for synthesizing a hologram of three-dimensional objects from multiple orthographic projection view images is proposed. The three-dimensional objects are captured under incoherent white illumination and their orthographic projection view images are obtained. The orthographic projection view images are multiplied by the corresponding phase terms and integrated to form a Fourier or Fresnel hologram. Using simple manipulation of the orthographic projection view images, it is also possible to shift the three-dimensional objects by an arbitrary amount along the three axes in the reconstruction space or invert their depths with respect to the given depth plane. The principle is verified experimentally.

Hologram Generation of 3D Objects Using Multiple Orthographic View Images

We propose a new synthesis method for the hologram of 3D objects using incoherent multiple orthographic view images. The 3D objects are captured and their multiple orthographic view images are generated from the captured image. Each orthographic view image is numerically overridden by the plane wave propagating in the direction of the corresponding view angle and integrated to form a point in the hologram plane. By repeating this process for all orthographic view images, we can generate the Fourier hologram of the 3D objects.