Real-time precise detection of regular grids and matrix codes

The traditional approach in detecting sets of concurrent and/or parallel lines is to first detect lines in the image and then find such groups of them which meet the concurrence condition. The Hough Transform can be used for detecting the lines and variants of HT such as the Cascaded Hough Transform can be used to detect the vanishing points. However, these approaches disregard much of the information actually accumulated to the Hough space. This article proposes using the Hough space as a 2D signal instead of just detecting the local maxima and processing them. On the example of QRcode detection, it is shown that this approach is computationally cheap, robust, and accurate. The proposed algorithm can be used for efficient and accurate detection and localization of matrix codes (QRcode, Aztec, DataMatrix, etc.) and chessboard-like calibration patterns.     

Real-time PCA calculation for spectral imaging (using SIMD and GP-GPU)

This article presents two optimized implementations of the PCA algorithm, primarily targeted on spectral image analysis in real time. One of them utilizes the SSE instruction set of contemporary CPUs, and the other one runs on graphics processors, using the CUDA environment. The implementations are evaluated and compared with a multithreaded C implementation compiled by an optimizing compiler and the results show speed-ups of around 10?×?which allows for using PCA on RGB and spectral images in real time. The discussed implementations are made available in a dynamically linked library, including a MATLAB plug-in interface so that they can be used by the professional public.     

Fast bilateral filter for HDR imaging

Bilateral filtering is a method often used in image processing applications. It is specifically useful for HDR algorithms. A novel approach to a fast and close approximation of bilateral filtering is presented. The method is designed especially with a focus on HDR image conversion into a normal color space processing. This paper presents the methods itself, describes the sources of acceleration and discusses the results of the method.     

EnMS: early non-maxima suppression

Detection of objects in images using statistical classifiers is a well studied and documented technique. Different applications of such detectors often require selection of the image position with the highest response of the detector—they perform non-maxima suppression. This article introduces the concept of early non-maxima suppression, which aims to reduce necessary computations by making the non-maxima suppression decision early based on incomplete information provided by a partially evaluated classifier. We show that the error of one such speculative decision with respect to a decision made based on response of the complete classifier can be estimated by collecting statistics on unlabeled data. The article then considers a sequential strategy of multiple early non-maxima suppression tests which follows the structure of soft-cascade detectors commonly used for object detection. We also show that an optimal (fastest for requested error rate) suppression strategy can be created by a novel variant of Wald’s sequential probability ratio test (SPRT) which we call the conditioned SPRT (CSPRT). Experimental results show that the early non-maxima suppression significantly reduces amount of computation in the case of object localization while the error rates are limited to low predefined values. The proposed approach notably outperforms the state-of-the-art detectors based on WaldBoost. The potential applications of the early non-maxima suppression approach are not limited to object localization and could be applied wherever the goal is to find the strongest response of a classifier among a set of classified samples.     

Real-time object detection on CUDA

The aim of the research described in this article is to accelerate object detection in images and video sequences using graphics processors. It includes algorithmic modifications and adjustments of existing detectors, constructing variants of efficient implementations and evaluation comparing with efficient implementations on the CPUs. This article focuses on detection by statistical classifiers based on boosting. The implementation and the necessary algorithmic alterations are described, followed by experimental measurements of the created object detector and discussion of the results. The final solution outperforms the reference efficient CPU/SSE implementation, by approximately 6–8× for high-resolution videos using nVidia GeForce 9800GTX and Intel Core2 Duo E8200.     

Real-time detection of lines using parallel coordinates and CUDA

The Hough transform is a well-known and popular algorithm for detecting lines in raster images. The standard Hough transform is rather slow to be usable in real time, so different accelerated and approximated algorithms exist. This study proposes a modified accumulation scheme for the Hough transform, using a new parameterization of lines “PClines”. This algorithm is suitable for computer systems with a small but fast read-write memory, such as today’s graphics processors. The algorithm requires no floating-point computations or goniometric functions. This makes it suitable for special and low-power processors and special-purpose chips. The proposed algorithm is evaluated both on synthetic binary images and on complex real-world photos of high resolutions. The results show that using today’s commodity graphics chips, the Hough transform can be computed at interactive frame rates, even with a high resolution of the Hough space and with the Hough transform fully computed.     

How to shoot yourself right with a smartphone?

Virtual Reality - This article presents a new use case of using handheld augmented reality to set up a smartphone camera for taking a self-photograph needed for evaluating the user’s sports...