José María González-Linares

Bidimensional shape detection using an invariant approach

Abstract Bidimensional shape detection is a process with high computational complexity. In this work, an algorithm, based on the generalized Hough transform (GHT), is presented in order to calculate the orientation, scale, and displacement of a image shape with respect to a template. To reduce the complexity, the uncoupled of the parameter calculation is carried out. The generation of the invariant information needed by the uncoupled is implemented by using three transformation functions that pair shape edge points. Differences between gradient vector angles are used to choose the paired points. An “a priori” study of the template shape is carried out to select the most suitable values for the difference angles.

A Clustering Technique for Video Copy Detection

In this work, a new method for detecting copies of a query video in a videos database is proposed. It includes a new clustering technique that groups frames with similar visual content, maintaining their temporal order. Applying this technique, a keyframe is extracted for each cluster of the query video. Keyframe choice is carried out by selecting the frame in the cluster with maximum similarity to the rest of frames in the cluster. Then, keyframes are compared to target videos frames in order to extract similarity regions in the target video. Relaxed temporal constraints are subsequently applied to the calculated regions in order to identify the copy sequence. The reliability and performance of the method has been tested by using several videos from the MPEG-7 Content Set, encoded with different frame sizes, bit rates and frame rates. Results show that our method obtains a significant improvement with respect to previous approaches in both achieved precision and computation time.

Logotype detection to support semantic-based video annotation

In conventional video production, logotypes are used to convey information about content originator or the actual video content. Logotypes contain information that is critical to infer genre, class and other important semantic features of video. This paper presents a framework to support semantic-based video classification and annotation. The backbone of the proposed framework is a technique for logotype extraction and recognition. The method consists of two main processing stages. The first stage performs temporal and spatial segmentation by calculating the minimal luminance variance region (MVLR) for a set of frames. Non-linear diffusion filters (NLDF) are used at this stage to reduce noise in the shape of the logotype. In the second stage, logotype classification and recognition are achieved. The earth movers distance (EMD) is used as a metric to decide if the detected MLVR belongs to one of the following logotype categories: learned or candidate. Learned logos are semantically annotated shapes available in the database. The semantic characterization of such logos is obtained through an iterative learning process. Candidate logos are non-annotated shapes extracted during the first processing stage. They are assigned to clusters grouping different instances of logos of similar shape. Using these clusters, false logotypes are removed and different instances of the same logo are averaged to obtain a unique prototype representing the underlying noisy cluster. Experiments involving several hours of MPEG video and around 1000 of candidate logotypes have been carried out in order to show the robustness of both detection and classification processes.

Performance models for asynchronous data transfers on consumer Graphics Processing Units

Graphics Processing Units (GPU) have impressively arisen as general-purpose coprocessors in high performance computing applications, since the launch of the Compute Unified Device Architecture (CUDA). However, they present an inherent performance bottleneck in the fact that communication between two separate address spaces (the main memory of the CPU and the memory of the GPU) is unavoidable. The CUDA Application Programming Interface (API) provides asynchronous transfers and streams, which permit a staged execution, as a way to overlap communication and computation. Nevertheless, a precise manner to estimate the possible improvement due to overlapping does not exist, neither a rule to determine the optimal number of stages or streams in which computation should be divided. In this work, we present a methodology that is applied to model the performance of asynchronous data transfers of CUDA streams on different GPU architectures. Thus, we illustrate this methodology by deriving expressions of performance for two different consumer graphic architectures belonging to the more recent generations. These models permit programmers to estimate the optimal number of streams in which the computation on the GPU should be broken up, in order to obtain the highest performance improvements. Finally, we have checked the suitability of our performance models with three applications based on codes from the CUDA Software Development Kit (SDK) with successful results.

An optimized approach to histogram computation on GPU

A histogram is a compact representation of the distribution of data in an image with a full range of applications in diverse fields. Histogram generation is an inherently sequential operation where every pixel votes in a reduced set of bins. This makes finding efficient parallel implementations very desirable but challenging, because on graphics processing units thousands of threads may be atomically updating a short number of histogram bins. Under these circumstances, collisions among threads will be very frequent and such collisions will serialize thread execution, seriously damaging the performance. In this paper we propose a highly optimized approach to histogram calculation, which tackles such performance bottlenecks. It uses histogram replication for eliminating position conflicts, padding to reduce bank conflicts, and an improved access to input data called interleaved read access. Our so-called

In-place transposition of rectangular matrices on accelerators

{\mathcal{R}}

Parallelization of a Video Segmentation Algorithm on CUDA---Enabled Graphics Processing Units

-per-block approach to histogram calculation has been successfully compared to the main state-of-the-art works using four histogram-based image processing kernels and two real image databases. Results show that our proposal is between 1.4 and 15.7 faster than every previous implementation for histograms of up to 4,096 bins.

A TV-logo classification and learning system

GPU application implementations using scatter approaches will fall into write contention due to atomic updates of output elements, if these result from more than one input element. Colliding threads will be serialized, seriously harming performance. Dealing with these issues requires a proper understanding of the behavior of the scratchpad or shared memory under conflicting accesses caused by concurrent threads. Thus, this paper presents an exhaustive microbenchmark-based analysis of atomic additions in shared memory that quantifies the impact of access conflicts on latency and throughput. This analysis has led us to discover the lock mechanism that enables atomic updates to shared memory and to propose a performance model to estimate the latency penalties due to collisions by position or bank conflicts. Then, we have derived experiments from this model that show us the way to optimize applications using atomic operations. Position and bank conflicts can be diminished by replication and padding, respectively. The benefits of such techniques are illustrated with the optimization of two widely used voting processes: the centroid updating step in k-means clustering, and histogram calculation.