Fernanda A. Andaló

Shape feature extraction and description based on tensor scale

Tensor scale is a morphometric parameter that unifies the representation of local structure thickness, orientation, and anisotropy, which can be used in several computer vision and image processing tasks. In this article, we exploit this concept for binary images and propose a shape salience detector and a shape descriptor-Tensor Scale Descriptor with Influence Zones. It also introduces a robust method to compute tensor scale, using a graph-based approach-the Image Foresting Transform. Experimental results are provided, showing the effectiveness of the proposed methods, when compared to other relevant methods, such as Beam Angle Statistics and Contour Salience Descriptor, with regard to their use in content-based image retrieval tasks.

Vanishing point detection by segment clustering on the projective space

The analysis of vanishing points on digital images provides strong cues for inferring the 3D structure of the depicted scene and can be exploited in a variety of computer vision applications. In this paper, we propose a method for estimating vanishing points in images of architectural environments that can be used for camera calibration and pose estimation, important tasks in large-scale 3D reconstruction. Our method performs automatic segment clustering in projective space --- a direct transformation from the image space --- instead of the traditional bounded accumulator space. Since it works in projective space, it handles finite and infinite vanishing points, without any special condition or threshold tuning. Experiments on real images show the effectiveness of the proposed method. We identify three orthogonal vanishing points and compute the estimation error based on their relation with the Image of the Absolute Conic (IAC) and based on the computation of the camera focal length.

Detecting Contour Saliences using Tensor Scale

Tensor scale is a morphometric parameter that unifies the representation of local structure thickness, orientation, and anisotropy, which can be used in several image processing tasks. This paper introduces a new application for tensor scale, which is the detection of saliences on a given contour, based on the tensor scale orientations computed for the entire object and mapped to its contour. For validation purposes, we present a shape descriptor that uses the detected contour saliences. Experimental results are provided, comparing the proposed method with our previous contour salience descriptor (CS). We show that the proposed method can be not only faster and more robust in the detection of salience points than the CS method, but also more effective as a shape descriptor.

PSQP: Puzzle Solving by Quadratic Programming

In this article we present the first effective method based on global optimization for the reconstruction of image puzzles comprising rectangle pieces—Puzzle Solving by Quadratic Programming (PSQP). The proposed novel mathematical formulation reduces the problem to the maximization of a constrained quadratic function, which is solved via a gradient ascent approach. The proposed method is deterministic and can deal with arbitrary identical rectangular pieces. We provide experimental results showing its effectiveness when compared to state-of-the-art approaches. Although the method was developed to solve image puzzles, we also show how to apply it to the reconstruction of simulated strip-shredded documents, broadening its applicability.In this article we present the first effective method based on global optimization for the reconstruction of image puzzles comprising rectangle pieces-Puzzle Solving by Quadratic Programming (PSQP). The proposed novel mathematical formulation reduces the problem to the maximization of a constrained quadratic function, which is solved via a gradient ascent approach. The proposed method is deterministic and can deal with arbitrary identical rectangular pieces. We provide experimental results showing its effectiveness when compared to state-of-the-art approaches. Although the method was developed to solve image puzzles, we also show how to apply it to the reconstruction of simulated strip-shredded documents, broadening its applicability.

Efficient height measurements in single images based on the detection of vanishing points

We propose a detector for identifying vanishing points in single images.We apply the detector to the problem of computing heights in single images.The height of a person was measured in several images.The mean observed error was 0.58?cm. Surveillance cameras have become a customary security equipment in buildings and streets worldwide. It is up to the field of Computational Forensics to provide automated methods for extracting and analyzing relevant image data captured by such equipment. In this article, we describe an effective and semi-automated method for detecting vanishing points, with their subsequent application to the problem of computing heights in single images. With no necessary camera calibration, our method iteratively clusters segments in the bi-dimensional projective space, identifying all vanishing points - finite and infinite - in an image. We conduct experiments on images of man-made environments to evaluate the output of the proposed method and we also consider its application on a photogrammetry framework.

Detecting vanishing points by segment clustering on the projective plane for single-view photogrammetry

In this paper, we describe how an effective vanishing point detector can be applied to photogrammetry when only a single view of an architectural environment is available. Our method performs automatic segment clustering in projective space — a direct transformation from the image space — instead of the traditional bounded accumulator space. Experiments on real images show the effectiveness of the proposed detector in finding all vanishing points, as well as its application in a photogrammetry algorithm, by recovering the vertical direction of the scene and the vanishing line for the ground plane.

Transmitting What Matters: Task-Oriented Video Composition and Compression

We present a simple yet effective framework – Transmitting What Matters (TWM) – to generate compressed videos containing only relevant objects targeted to specific computer vision tasks, such as faces for the task of face expression recognition, license plates for the task of optical character recognition, among others. TWM takes advantage of the final desired computer vision task to compose video frames only with the necessary data. The video frames are compressed and can be stored or transmitted to powerful servers where extensive and time-consuming tasks can be performed. We experimentally present the trade-offs between distortion and bitrate for a wide range of compression levels, and the impact generated by compression artifacts on the accuracy of the desired vision task. We show that, for one selected computer vision task, it is possible to dramatically reduce the amount of required data to be stored or transmitted, without compromising accuracy.

CrowdPet: Deep learning applied to the detection of dogs in the wild

Biometry is the statistical study of physical or behavioral characteristics of living beings, mainly applied to the identification of individuals. Visual biometric methods map images to identities, and have been applied almost exclusively to humans. However, animal visual biometry is significantly different than that of humans, and current methods cannot be used interchangeably. The CrowdPet initiative studies and proposes methods to the problem of animal biometric identification, in order to identify and generate analytics related to stray animals. As part of this effort, in this work we present a method for detecting dogs in images, despite changes in pose and illumination, using Deep Learning. Experiments show promising results, in terms of accuracy, using challenging current datasets.

TWM: A framework for creating highly compressible videos targeted to computer vision tasks

Abstract We present a simple yet effective framework – Transmitting What Matters (TWM) – to generate highly compressible videos containing only relevant information targeted to specific computer vision tasks, such as faces for the task of face expression recognition, license plates for the task of optical character recognition, among others. TWM takes advantage of the final desired computer vision task to compose video frames only with the necessary data. The video frames are compressed and can be stored or transmitted to powerful servers where extensive and time-consuming tasks are performed. Experiments explore the trade-offs between distortion and bitrate for a wide range of compression levels, and the impact generated by compression artifacts on the accuracy of the desired vision task. We show that, for two computer vision tasks implemented by different methods, it is possible to dramatically reduce the amount of required data to be stored or transmitted, without compromising accuracy. With PSNR YUV quality of over 41  dB, the bitrate was reduced up to four times, while a detection task was affected by only  ∼ 1 pixel and a classification task by 1 ∼ 2 percentage points.