Search

The widening of the complexity-productivity gap in application development witnessed in the last years is becoming an important issue for the developers. New design methods try to automate most designers tasks to bridge this gap. In addition, new Model of Computations (MoCs), as those dataflow-based, ease the expression of parallelism within applications, leading to higher designer productivity. Rapid prototyping design tools offer fast estimations of the soundness of design choices. A key step when prototyping an application is to have representative performance indicators to estimate the validity of those design choices. Such indicators can be obtained using hardware information, while new libraries, e.g.,Performance Application Programming Interface (PAPI), ease the access to such hardware information. In this work, Papify toolbox is presented as a tool to perform automatic PAPI-based instrumentation of dynamic dataflow applications. It combines Papify with a dataflow Y-chart based design framework, which is called Preesm, and its companion run-time reconfiguration manager, which is called Synchronous Parameterized and Interfaced Dataflow Embedded Runtime (SPiDER). Papify toolbox accounts for an automatic code generator for static and dynamic applications, a dedicated library to manage the monitoring at run-time and two User Interfaces (UIs) to ease both the configuration and the analysis of the captured run-time information. Additionally, its main advantages are 1) its capability of adapting the monitoring according to the system status and 2) adaptation of the monitoring accordingly to application workload redistribution in run-time. A thorough overhead characterization using Sobel-morpho and Stereo-matching dataflow applications shows that Papify run-time monitoring overhead is up to 10%.

The use of hyperspectral imaging for medical applications is becoming more common in recent years. One of the main obstacles that researchers find when developing hyperspectral algorithms for medical applications is the lack of specific, publicly available, and hyperspectral medical data. The work described in this paper was developed within the framework of the European project HELICoiD (HypErspectraL Imaging Cancer Detection), which had as a main goal the application of hyperspectral imaging to the delineation of brain tumors in real-time during neurosurgical operations. In this paper, the methodology followed to generate the first hyperspectral database of in-vivo human brain tissues is presented. Data was acquired employing a customized hyperspectral acquisition system capable of capturing information in the Visual and Near InfraRed (VNIR) range from 400 to 1000 nm. Repeatability was assessed for the cases where two images of the same scene were captured consecutively. The analysis reveals that the system works more efficiently in the spectral range between 450 and 900 nm. A total of 36 hyperspectral images from 22 different patients were obtained. From these data, more than 300 000 spectral signatures were labeled employing a semi-automatic methodology based on the spectral angle mapper algorithm. Four different classes were defined: normal tissue, tumor tissue, blood vessel, and background elements. All the hyperspectral data has been made available in a public repository.

Hyperspectral (HS) imaging presents itself as a non-contact, non-ionizing and non-invasive technique, proven to be suitable for medical diagnosis. However, the volume of information contained in these images makes difficult providing the surgeon with information about the boundaries in real-time. To that end, High-Performance-Computing (HPC) platforms become necessary. This paper presents a comparison between the performances provided by five different HPC platforms while processing a spatial-spectral approach to classify HS images, assessing their main benefits and drawbacks. To provide a complete study, two different medical applications, with two different requirements, have been analyzed. The first application consists of HS images taken from neurosurgical operations; the second one presents HS images taken from dermatological interventions. While the main constraint for neurosurgical applications is the processing time, in other environments, as the dermatological one, other requirements can be considered. In that sense, energy efficiency is becoming a major challenge, since this kind of applications are usually developed as hand-held devices, thus depending on the battery capacity. These requirements have been considered to choose the target platforms: on the one hand, three of the most powerful Graphic Processing Units (GPUs) available in the market; and, on the other hand, a low-power GPU and a manycore architecture, both specifically thought for being used in battery-dependent environments.

This paper presents a study of the adaptation of a Non-Linear Iterative Partial Least Squares (NIPALS) algorithm applied to Hyperspectral Imaging to a Massively Parallel Processor Array manycore architecture, which assembles 256 cores distributed over 16 clusters. This work aims at optimizing the internal communications of the platform to achieve real-time processing of large data volumes with limited computational resources and memory bandwidth. As hyperspectral images are composed of extensive volumes of spectral information, real-time requirements, which are upper-bounded by the image capture rate of the hyperspectral sensor, are a challenging objective. To address this issue, the image size is usually reduced prior to the processing phase, which is itself a computationally intensive task. Consequently, this paper proposes an analysis of the intrinsic parallelism and the data dependency within the NIPALS algorithm and its subsequent implementation on a manycore architecture. Furthermore, this implementation has been validated against three hyperspectral images extracted from both remote sensing and medical datasets. As a result, an average speedup of 17× has been achieved when compared to the sequential version. Finally, this approach has been compared with other state-of-the-art implementations, outperforming them in terms of performance.

This paper presents a study of the design space of a Support Vector Machine (SVM) classifier with a linear kernel running on a manycore MPPA (Massively Parallel Processor Array) platform. This architecture gathers 256 cores distributed in 16 clusters working in parallel. This study aims at implementing a real-time hyperspectral SVM classifier, where real-time is defined as the time required to capture a hyperspectral image. To do so, two aspects of the SVM classifier have been analyzed: the classification algorithm and the system parallelization. On the one hand, concerning the classification algorithm, first, the classification model has been optimized to fit into the MPPA structure and, secondly, a probability estimation stage has been included to refine the classification results. On the other hand, the system parallelization has been divided into two levels: first, the parallelism of the classification has been exploited taking advantage of the pixel-wise classification methodology supported by the SVM algorithm and, secondly, a double-buffer communication procedure has been implemented to parallelize the image transmission and the cluster classification stages. Experimenting with medical images, an average speedup of 9 has been obtained using a single-cluster and double-buffer implementation with 16 cores working in parallel. As a result, a system whose processing time linearly grows with the number of pixels composing the scene has been implemented. Specifically, only 3 µs are required to process each pixel within the captured scene independently from the spatial resolution of the image.

International audience; In this paper, we propose a runtime framework that implements code multi-versioning and specialization to optimize and parallelize loop kernels that are invoked many times with varying parameters. These parameters may influence the code structure, the touched memory locations, the work-load, and the runtime performance. They may also impact the validity of the parallelizing and optimizing polyhedral transformations that are applied on-the-fly. For a target loop kernel and its associated parameters, a different optimizing and parallelizing transformation is evaluated at each invocation, among a finite set of transformations (multi-versioning and specialization). The best performing transformed code version is stored and indexed using its associated parameters. When every optimizing transformation has been evaluated, the best performing code version regarding the current parameters, which has been stored, is relaunched at next invocations (memoization).

Cyber-Physical Systems (CPS) are interconnected devices, reactive and dynamic to sensed external and internal triggers. The H2020 CERBERO EU Project is developing a design environment composed by modelling, deployment and verification tools for adaptive CPS. This paper focuses on its efficient support for run-time self-adaptivity.

In the era of Cyber Physical Systems, designers need to offer support for run-time adaptivity considering different constraints, including the internal status of the system. This work proposes a run-time monitoring approach for hardware accelerators, based on the Performance Application Programming Interface.

In this work, a standard and unified method for monitoring hardware accelerators in Reconfigurable Computing Architectures is proposed, based on a standard software monitoring interface. The open source Performance Application Programming Interface (PAPI) library is commonly used in the field of High Performance Computing and aims at providing event information directly extracted from a set of Performance Monitor Counters. Important events such as data and instruction cache misses, hardware interrupts, etc. are collected to analyze and profile applications to pinpoint the contingent bottlenecks. In other words, it serves as a ”Hardware Abstraction layer” for applications running in the user-space. In this paper, its use is extended by proposing a method to target custom Performance Hardware Registers on accelerators built upon an FPGA. Furthermore, portability and standardization are discussed and the overhead associated with PAPI use is evaluated. Two hardware examples are proposed to evaluate this approach: a simple counter and an infrastructure for hardware accelerators called ARTICo3.

Dimensionality reduction represents a critical preprocessing step in order to increase the efficiency and the performance of many hyperspectral imaging algorithms. However, dimensionality reduction algorithms, such as the Principal Component Analysis (PCA), suffer from their computationally demanding nature, becoming advisable for their implementation onto high-performance computer architectures for applications under strict latency constraints. This work presents the implementation of the PCA algorithm onto two different high-performance devices, namely, an NVIDIA Graphics Processing Unit (GPU) and a Kalray manycore, uncovering a highly valuable set of tips and tricks in order to take full advantage of the inherent parallelism of these high-performance computing platforms, and hence, reducing the time that is required to process a given hyperspectral image. Moreover, the achieved results obtained with different hyperspectral images have been compared with the ones that were obtained with a field programmable gate array (FPGA)-based implementation of the PCA algorithm that has been recently published, providing, for the first time in the literature, a comprehensive analysis in order to highlight the pros and cons of each option.

Hyperspectral imaging (HSI) allows for the acquisition of large numbers of spectral bands throughout the electromagnetic spectrum (within and beyond the visual range) with respect to the surface of scenes captured by sensors. Using this information and a set of complex classification algorithms, it is possible to determine which material or substance is located in each pixel. The work presented in this paper aims to exploit the characteristics of HSI to develop a demonstrator capable of delineating tumor tissue from brain tissue during neurosurgical operations. Improved delineation of tumor boundaries is expected to improve the results of surgery. The developed demonstrator is composed of two hyperspectral cameras covering a spectral range of 400-1700 nm. Furthermore, a hardware accelerator connected to a control unit is used to speed up the hyperspectral brain cancer detection algorithm to achieve processing during the time of surgery. A labeled dataset comprised of more than 300,000 spectral signatures is used as the training dataset for the supervised stage of the classification algorithm. In this preliminary study, thematic maps obtained from a validation database of seven hyperspectral images of in vivo brain tissue captured and processed during neurosurgical operations demonstrate that the system is able to discriminate between normal and tumor tissue in the brain. The results can be provided during the surgical procedure (~1 min), making it a practical system for neurosurgeons to use in the near future to improve excision and potentially improve patient outcomes.

International audience; The widening of the complexity-productivity gap witnessed in the last years is becoming unaffordable from the application development point of view. New design methods try to automate most designers tasks in order to bridge this gap. In addition, new Models of Computation (MoC), as those dataflow-based, ease the expression of parallelism within applications and lead to higher productivity. Rapid prototyping design tools offer fast estimations of the soundness of design choices. A key step when prototyping an application is to have representative performance indicators to estimate the validity of the design choices. Such indicators can be obtained using hardware information through the Performance API (PAPI). In this work, PAPI and a dataflow MoC are integrated within a Y-chart design flow. The implementation takes the form of a dedicated automatic code generation scheme within the Preesm tool. Preliminary results show that depending on the complexity of the application, the computation time overhead due to monitoring varies from being almost negligible to more than 50%. Also, on top of offering accurate hardware performance indicators, the extracted values can be combined to estimate power or energy consumption. © 2018 Association for Computing Machinery.

This paper presents a study of the par alle lization possibilities of a Non-Linear Iterative Partial Least Squares algorithm and its adaptation to a Massively Parallel Processor Array manycore architecture, which assembles 256 cores distributed over 16 clusters. The aim of this work is twofold: first, to test the behavior of iterative, complex algorithms in a manycore architecture; and, secondly, to achieve real-time processing of hyperspectral images, which is fixed by the image capture rate of the hyperspectral sensor. Real-time is a challenging objective, as hyperspectral images are composed of extensive volumes of spectral information. This issue is usually addressed by reducing the image size prior to the processing phase itself. Consequently, this paper proposes an analysis of the intrinsic parallelism of the algorithm and its subsequent implementation on a manycore architecture. As a result, an average speedup of 13 has been achieved when compared to the sequential version. Additionally, this implementation has been compared with other state-of-the-art applications, outperforming them in terms of performance.

In this paper, a Massively Parallel Processor Array platform is characterized in terms of energy consumption using a Support Vector Machine for hyperspectral image classification. This platform gathers 16 clusters composed of 16 cores each, i.e., 256 processors working in parallel. The objective of the work is to associate power dissipation and energy consumed by the platform with the different resources of the architecture. Experimenting with a hyperspectral SVM classifier, this study has been conducted using three strategies: i) modifying the number of processing elements, i.e., clusters and cores, ii) increasing system frequency, and iii) varying the number of active communication links during the analysis, i.e., I/Os and DMAs. As a result, a relationship between the energy consumption and the active platform resources has been exposed using two different parallelization strategies. Finally, the implementation that fully exploits the parallelization possibilities working at 500MHz has been proven to be also the most efficient one, as it reduces the energy consumption by 98% when compared to the sequential version running at 400MHz.

Current computational demands require increasing designer's efficiency and system performance per watt. A broadly accepted solution for efficient accelerators implementation is reconfigurable computing. However, typical HDL methodologies require very specific skills and a considerable amount of designer's time. Despite the new approaches to high-level synthesis like OpenCL, given the large heterogeneity in today's devices (manycore, CPUs, GPUs, FPGAs), there is no one-fits-all solution, so to maximize performance, platform-driven optimization is needed. This paper reviews some latest works using Intel FPGA SDK for OpenCL and the strategies for optimization, evaluating the framework for the design of a hyperspectral image spatial-spectral classifier accelerator. Results are reported for a Cyclone V SoC using Intel FPGA OpenCL Offline Compiler 16.0 out-of-the-box. From a common baseline C implementation running on the embedded ARM® Cortex®-A9, OpenCL-based synthesis is evaluated applying different generic and vendor specific optimizations. Results show how reasonable speedups are obtained in a device with scarce computing and embedded memory resources. It seems a great step has been given to effectively raise the abstraction level, but still, a considerable amount of HW design skills is needed.

The HELICoiD project is a European FP7 FET Open funded project. It is an interdisciplinary work at the edge of the biomedical domain, bringing together neurosurgeons, computer scientists and electronic engineers. The main target of the project was to provide a working demonstrator of an intraoperative image-guided surgery system for real-time brain cancer detection, in order to assist neurosurgeons during tumour resection procedures. One of the main problems associated to brain tumours is its infiltrative nature, which makes complete tumour resection a highly difficult task. With the combination of Hyperspectral Imaging and Machine Learning techniques, the project aimed at demonstrating that a precise determination of tumour boundaries was possible, helping this way neurosurgeons to minimize the amount of removed healthy tissue. The project partners involved, besides different universities and companies, two hospitals where the demonstrator was tested during surgical procedures. This paper introduces the difficulties around brain tumor resection, stating the main objectives of the project and presenting the materials, methodologies and platforms used to propose a solution. A brief summary of the main results obtained is also included.

Introduction: Hyperspectral imaging (HSI) is an emerging technology for medical diagnosis. In this research work, a multidisciplinary team, made up of pathologists and engineers, presents a proof of concept on the use of HSI analysis in order to automatically detect human brain tumour tissue from pathological slides. The samples were acquired from four different patients diagnosed with high-grade gliomas. Based on the diagnosis provided by pathologists, a spectral library containing spectra from healthy and tumour tissues was created. Data were finally processed using three different supervised machine learning algorithms. Materials and methods: An acquisition system consisting of a HSI camera coupled to a microscope was developed to capture the hyperspectral images from pathology slides. The spectral sampling was done in the spectral range from 400 nm to 1000 nm with a spectral resolution of 2.8 nm. The biological samples consisted of biopsies of human brain tissue resected during surgery that followed a histological process, whereby tissue specimens were prepared for sectioning, staining and diagnosis. Only spectral characteristics of the data were taken into account. The inputs of the classifiers were the spectral signatures from healthy and tumour pixels. Three different supervised machine learning algorithms were employed: Support Vector Machines (SVM), Artificial Neural Networks (ANN) and Random Forests (RF). Results: The automatic diagnosis provided by the supervised classifiers shows a very high discrimination rate between healthy and tumour tissue, with high specificity and sensitivity above 90.83% and 94.55% respectively. Although all classifiers provide an accurate discrimination between healthy and tumour tissue, ANN presents the most accurate results with a specificity of 98.72% and a sensitivity of 97.71%. Conclusions: This research work presents a proof of concept in the use of HSI for automatically detecting brain tumour tissue in pathological slides. HSI can obtain an accurate diagnosis without using the morphological features of tissues, being a suitable complement to the current analysis methods, assisting pathologists to analyse the slides without having to spend a long time in the examination of each sample.FUNDING: This work has been supported by the European Commission through the FP7 FET Open programme ICT- 2011.9.2, European Project HELICoiD “HypErspectral Imaging Cancer Detection” under Grant Agreement 618080.

International audience; This paper presents a study of the parallelism of a Principal Component Analysis (PCA) algorithm and its adaptation to a manycore MPPA (Massively Parallel Processor Array) architecture, which gathers 256 cores distributed among 16 clusters. This study focuses on porting hyperspectral image processing into many core platforms by optimizing their processing to fulfill real-time constraints, fixed by the image capture rate of the hyperspectral sensor. Real-time is a challenging objective for hyperspectral image processing, as hyperspectral images consist of extremely large volumes of data and this problem is often solved by reducing image size before starting the processing itself. To tackle the challenge, this paper proposes an analysis of the intrinsic parallelism of the different stages of the PCA algorithm with the objective of exploiting the parallelization possibilities offered by an MPPA manycore architecture. Furthermore, the impact on internal communication when increasing the level of parallelism, is also analyzed. Experimenting with medical images obtained from two different surgical use cases, an average speedup of 20 is achieved. Internal communications are shown to rapidly become the bottleneck that reduces the achievable speedup offered by the PCA parallelization. As a result of this study, PCA processing time is reduced to less than 6 s, a time compatible with the targeted brain surgery application requiring 1 frame-per-minute.

Hyperspectral Imaging (HI) collects high resolution spectral information consisting of hundreds of bands across the electromagnetic spectrum –from the ultraviolet to the infrared range–. Thanks to this huge amount of information, an identification of the different elements that compound the hyperspectral image is feasible. Initially, HI was developed for remote sensing applications and, nowadays, its use has been spread to research fields such as security and medicine. In all of them, new applications that demand the specific requirement of real-time processing have appear. In order to fulfill this requirement, the intrinsic parallelism of the algorithms needs to be explicitly exploited. In this paper, a Support Vector Machine (SVM) classifier with a linear kernel has been implemented using a dataflow language called RVC-CAL. Specifically, RVC-CAL allows the scheduling of functional actors onto the target platform cores. Once the parallelism of the classifier has been extracted, a comparison of the SVM classifier implementation using LibSVM –a specific library for SVM applications– and RVC-CAL has been performed. The speedup results obtained for the image classifier depends on the number of blocks in which the image is divided; concretely, when 3 image blocks are processed in parallel, an average speed up above 2.50, with regard to the RVC-CAL sequential version, is achieved.

In this paper, a study of the parallel exploitation of a Support Vector Machine (SVM) classifier with a linear kernel running on a Massively Parallel Processor Array platform is exposed. This system joins 256 cores working in parallel and grouped in 16 different clusters. The main objective of the research has been to develop an optimal implementation of the SVM classifier on a MPPA platform whilst the architectural bottlenecks of the hyperspectral image classifier are analyzed. Experimenting with medical images, the parallelization of the SVM classification has been conducted using three strategies: i) single- and multi-core processing, ii) single- and multi-cluster analysis and iii) single- and double-buffer execution. As a result, an average core processing speedup of 11.8 has been achieved when parallelizing the SVM classification process in a single cluster. On the contrary, since data communication accounts for 34.7% of the total execution time in the sequential case, the total speedup is upper-bounded to 2.9. Using a double-buffer methodology, a total speedup of 2.84 has been achieved on a single cluster. At last, the feasibility of a portable version of a linear SVM has been demonstrated.

In this paper, a demonstrator of three different elements of the EU FET HELICoiD project is introduced. The goal of this demonstration is to show how the combination of hyperspectral imaging and machine learning can be a potential solution to precise real-time detection of tumor tissues during surgical operations. The HELICoiD setup consists of two hyperspectral cameras, a scanning unit, an illumination system, a data processing system and an EMB01 accelerator platform, which hosts an MPPA-256 manycore chip. All the components are mounted fulfilling restrictions from surgical environments, as shown in the accompanying video recorded at the operating room. An in-vivo human brain hyperspectral image data base, obtained at the University Hospital Doctor Negrin in Las Palmas de Gran Canaria, has been employed as input to different supervised classification algorithms (SVM, RF, NN) and to a spatial-spectral filtering stage (SVM-KNN). The resulting classification maps are shown in this demo. In addition, the implementation of the SVM-KNN classification algorithm on the MPPA EMB01 platform is demonstrated in the live demo.

International audience; Hyperspectral imaging (HI) collects information from across the electromagnetic spectrum, covering a wide range of wavelengths. The tremendous development of this technology within the field of remote sensing has led to new research fields, such as cancer automatic detection or precision agriculture, but has also increased the performance requirements of the applications. For instance, strong time constraints need to be respected, since many applications imply real-time responses. Achieving real-time is a challenge, as hyperspectral sensors generate high volumes of data to process. Thus, so as to achieve this requisite, first the initial image data needs to be reduced by discarding redundancies and keeping only useful information. Then, the intrinsic parallelism in a system specification must be explicitly highlighted. In this paper, the PCA (Principal Component Analysis) algorithm is implemented using the RVC-CAL dataflow language, which specifies a system as a set of blocks or actors and allows its parallelization by scheduling the blocks over different processing units. Two implementations of PCA for hyperspectral images have been compared when aiming at obtaining the first few principal components: first, the algorithm has been implemented using the Jacobi approach for obtaining the eigenvectors; thereafter, the NIPALS-PCA algorithm, which approximates the principal components iteratively, has also been studied. Both implementations have been compared in terms of accuracy and computation time; then, the parallelization of both models has also been analyzed. These comparisons show promising results in terms of computation time and parallelization: the performance of the NIPALS-PCA algorithm is clearly better when only the first principal component is achieved, while the partitioning of the algorithm execution over several cores shows an important speedup for the PCA-Jacobi. Thus, experimental results show the potential of RVC-CAL to automatically generate implementations which process in real-time the large volumes of information of hyperspectral sensors, as it provides advanced semantics for exploiting system parallelization. © 2016 SPIE.

The use of hyperspectral imaging (HSI) in the medical field is an emerging approach to assist physicians in diagnostic or surgical guidance tasks. However, HSI data processing involves very high computational requirements due to the huge amount of information captured by the sensors. One of the stages with higher computational load is the K-Nearest Neighbors (KNN) filtering algorithm. The main goal of this study is to optimize and parallelize the KNN algorithm by exploiting the GPU technology to obtain real-time processing during brain cancer surgical procedures. This parallel version of the KNN performs the neighbor filtering of a classification map (obtained from a supervised classifier), evaluating the different classes simultaneously. The undertaken optimizations and the computational capabilities of the GPU device throw a speedup up to 66.18&times, when compared to a sequential implementation.

Surgery for brain cancer is a major problem in neurosurgery. The diffuse infiltration into the surrounding normal brain by these tumors makes their accurate identification by the naked eye difficult. Since surgery is the common treatment for brain cancer, an accurate radical resection of the tumor leads to improved survival rates for patients. However, the identification of the tumor boundaries during surgery is challenging. Hyperspectral imaging is a non-contact, non-ionizing and non-invasive technique suitable for medical diagnosis. This study presents the development of a novel classification method taking into account the spatial and spectral characteristics of the hyperspectral images to help neurosurgeons to accurately determine the tumor boundaries in surgical-time during the resection, avoiding excessive excision of normal tissue or unintentionally leaving residual tumor. The algorithm proposed in this study to approach an efficient solution consists of a hybrid framework that combines both supervised and unsupervised machine learning methods. Firstly, a supervised pixel-wise classification using a Support Vector Machine classifier is performed. The generated classification map is spatially homogenized using a one-band representation of the HS cube, employing the Fixed Reference t-Stochastic Neighbors Embedding dimensional reduction algorithm, and performing a K-Nearest Neighbors filtering. The information generated by the supervised stage is combined with a segmentation map obtained via unsupervised clustering employing a Hierarchical K-Means algorithm. The fusion is performed using a majority voting approach that associates each cluster with a certain class. To evaluate the proposed approach, five hyperspectral images of surface of the brain affected by glioblastoma tumor in vivo from five different patients have been used. The final classification maps obtained have been analyzed and validated by specialists. These preliminary results are promising, obtaining an accurate delineation of the tumor area.

Hyperspectral Imaging (HI) collects high resolution spectral information consisting of hundred of bands raging from the infrared to the ultraviolet wave lengths. In the medical field, specifically, in the cancer tissue identification at the operating room, the potential of HI is huge. However, given the data volume of HI and the computational complexity and cost of identification algorithms, real-time processing is the key, differential feature that brings value to surgeons. In order to achieve real-time implementations, the parallelism available in a specification needs to be explicitly highlighted. Data-flow programming languages, like RVC-CAL, are able to accomplish this goal. In this paper, an RVC-CAL library to implement dimensionality reduction and endmember extraction is presented. The results obtained show significant improvements with regard to a state-of-the-art analysis tool. A speedup of 30% is carried out using the complete processing chain and, in particular, a speedup of 5% has been achieved in the dimensionality reduction step. This dimensionality reduction takes ten of the thirteen seconds that the whole system needs to analyze one of the images. In addition, the RVC-CAL library is an excellent tool to simplify the implementation process of HI algorithms. Effectively, during the experimental test, the potential of the RVC-CAL library to reveal possible bottlenecks present in the HI processing chain and, therefore, to improve the system performance to achieve real-time constraints has been shown. Furthermore, the RVC-CAL library provides the possibility of system performance testing.

Hyperspectral imaging (HI) collects information from across the electromagnetic spectrum, covering a wide range of wavelengths. Although this technology was initially developed for remote sensing and earth observation, its multiple advantages - such as high spectral resolution - led to its application in other fields, as cancer detection. However, this new field has shown specific requirements; for instance, it needs to accomplish strong time specifications, since all the potential applications - like surgical guidance or in vivo tumor detection - imply real-time requisites. Achieving this time requirements is a great challenge, as hyperspectral images generate extremely high volumes of data to process. Thus, some new research lines are studying new processing techniques, and the most relevant ones are related to system parallelization. In that line, this paper describes the construction of a new hyperspectral processing library for RVC–CAL language, which is specifically designed for multimedia applications and allows multithreading compilation and system parallelization. This paper presents the development of the required library functions to implement two of the four stages of the hyperspectral imaging processing chain--endmember and abundances estimation. The results obtained show that the library achieves speedups of 30%, approximately, comparing to an existing software of hyperspectral images analysis; concretely, the endmember estimation step reaches an average speedup of 27.6%, which saves almost 8 seconds in the execution time. It also shows the existence of some bottlenecks, as the communication interfaces among the different actors due to the volume of data to transfer. Finally, it is shown that the library considerably simplifies the implementation process. Thus, experimental results show the potential of a RVC–CAL library for analyzing hyperspectral images in real-time, as it provides enough resources to study the system performance.