PolyMage: Automatic Optimization for Image Processing Pipelines

This paper presents the design and implementation of PolyMage, a domain-specific language and compiler for image processing pipelines. An image processing pipeline can be viewed as a graph of interconnected stages which process images successively. Each stage typically performs one of point-wise, stencil, reduction or data-dependent operations on image pixels. Individual stages in a pipeline typically exhibit abundant data parallelism that can be exploited with relative ease. However, the stages also require high memory bandwidth preventing effective utilization of parallelism available on modern architectures. For applications that demand high performance, the traditional options are to use optimized libraries like OpenCV or to optimize manually. While using libraries precludes optimization across library routines, manual optimization accounting for both parallelism and locality is very tedious. The focus of our system, PolyMage, is on automatically generating high-performance implementations of image processing pipelines expressed in a high-level declarative language. Our optimization approach primarily relies on the transformation and code generation capabilities of the polyhedral compiler framework. To the best of our knowledge, this is the first model-driven compiler for image processing pipelines that performs complex fusion, tiling, and storage optimization automatically. Experimental results on a modern multicore system show that the performance achieved by our automatic approach is up to 1.81x better than that achieved through manual tuning in Halide, a state-of-the-art language and compiler for image processing pipelines. For a camera raw image processing pipeline, our performance is comparable to that of a hand-tuned implementation.

Dual-DSP system for signal and image processing

The design of a dual-DSP microprocessor system and its application for parallel FFT and two-dimensional convolution are explained. The system is based on a master-salve configuration. Two ADSP-2101s are configured as slave processors and a PC/AT serves as the master. The master serves as a control processor to transfer the program code and data to the DSPs. The system architecture and the algorithms for the two applications, viz. FFT and two-dimensional convolutions, are discussed.

A Perspective on Computer Assisted Assessment Techniques for Short Free-Text Answers

Computer Assisted Assessment (CAA) has been existing for several years now. While some forms of CAA do not require sophisticated text understanding (e.g., multiple choice questions), there are also student answers that consist of free text and require analysis of text in the answer. Research towards the latter till date has concentrated on two main sub-tasks: (i) grading of essays, which is done mainly by checking the style, correctness of grammar, and coherence of the essay and (ii) assessment of short free-text answers. In this paper, we present a structured view of relevant research in automated assessment techniques for short free-text answers. We review papers spanning the last 15 years of research with emphasis on recent papers. Our main objectives are two folds. First we present the survey in a structured way by segregating information on dataset, problem formulation, techniques, and evaluation measures. Second we present a discussion on some of the potential future directions in this domain which we hope would be helpful for researchers.

FAST AND ACCURATE BILATERAL FILTERING USING GAUSS-POLYNOMIAL DECOMPOSITION

The bilateral filter is a versatile non-linear filter that has found diverse applications in image processing, computer vision, computer graphics, and computational photography. A common form of the filter is the Gaussian bilateral filter in which both the spatial and range kernels are Gaussian. A direct implementation of this filter requires O(sigma(2)) operations per pixel, where sigma is the standard deviation of the spatial Gaussian. In this paper, we propose an accurate approximation algorithm that can cut down the computational complexity to O(1) per pixel for any arbitrary sigma (constant-time implementation). This is based on the observation that the range kernel operates via the translations of a fixed Gaussian over the range space, and that these translated Gaussians can be accurately approximated using the so-called Gauss-polynomials. The overall algorithm emerging from this approximation involves a series of spatial Gaussian filtering, which can be efficiently implemented (in parallel) using separability and recursion. We present some preliminary results to demonstrate that the proposed algorithm compares favorably with some of the existing fast algorithms in terms of speed and accuracy.

GAP-RBF based NR image quality measurement for JPEG coded images

In this paper, we present a growing and pruning radial basis function based no-reference (NR) image quality model for JPEG-coded images. The quality of the images are estimated without referring to their original images. The features for predicting the perceived image quality are extracted by considering key human visual sensitivity factors such as edge amplitude, edge length, background activity and background luminance. Image quality estimation involves computation of functional relationship between HVS features and subjective test scores. Here, the problem of quality estimation is transformed to a function approximation problem and solved using GAP-RBF network. GAP-RBF network uses sequential learning algorithm to approximate the functional relationship. The computational complexity and memory requirement are less in GAP-RBF algorithm compared to other batch learning algorithms. Also, the GAP-RBF algorithm finds a compact image quality model and does not require retraining when the new image samples are presented. Experimental results prove that the GAP-RBF image quality model does emulate the mean opinion score (MOS). The subjective test results of the proposed metric are compared with JPEG no-reference image quality index as well as full-reference structural similarity image quality index and it is observed to outperform both.

A New Iterative FIR Filter for Image and Video Restoration

Image and video filtering is a key image-processing task in computer vision especially in noisy environment. In most of the cases the noise source is unknown and hence possess a major difficulty in the filtering operation. In this paper we present an error-correction based learning approach for iterative filtering. A new FIR filter is designed in which the filter coefficients are updated based on Widrow-Hoff rule. Unlike the standard filter the proposed filter has the ability to remove noise without the a priori knowledge of the noise. Experimental result shows that the proposed filter efficiently removes the noise and preserves the edges in the image. We demonstrate the capability of the proposed algorithm by testing it on standard images infected by Gaussian noise and on a real time video containing inherent noise. Experimental result shows that the proposed filter is better than some of the existing standard filters

Demonstrating the Usefulness of CAELinux for Computer Aided Engineering using an Example of the Three Dimensional Reconstruction of a Pig Liver

CAELinux is a Linux distribution which is bundled with free software packages related to Computer Aided Engineering (CAE). The free software packages include software that can build a three dimensional solid model, programs that can mesh a geometry, software for carrying out Finite Element Analysis (FEA), programs that can carry out image processing etc. Present work has two goals: 1) To give a brief description of CAELinux 2) To demonstrate that CAELinux could be useful for Computer Aided Engineering, using an example of the three dimensional reconstruction of a pig liver from a stack of CT-scan images. One can note that instead of using CAELinux, using commercial software for reconstructing the liver would cost a lot of money. One can also note that CAELinux is a free and open source operating system and all software packages that are included in the operating system are also free. Hence one can conclude that CAELinux could be a very useful tool in application areas like surgical simulation which require three dimensional reconstructions of biological organs. Also, one can see that CAELinux could be a very useful tool for Computer Aided Engineering, in general.

Bilateral smoothing of gradient vector field and application to image segmentation

Medical image segmentation finds application in computer-aided diagnosis, computer-guided surgery, measuring tissue volumes, locating tumors, and pathologies. One approach to segmentation is to use active contours or snakes. Active contours start from an initialization (often manually specified) and are guided by image-dependent forces to the object boundary. Snakes may also be guided by gradient vector fields associated with an image. The first main result in this direction is that of Xu and Prince, who proposed the notion of gradient vector flow (GVF), which is computed iteratively. We propose a new formalism to compute the vector flow based on the notion of bilateral filtering of the gradient field associated with the edge map - we refer to it as the bilateral vector flow (BVF). The range kernel definition that we employ is different from the one employed in the standard Gaussian bilateral filter. The advantage of the BVF formalism is that smooth gradient vector flow fields with enhanced edge information can be computed noniteratively. The quality of image segmentation turned out to be on par with that obtained using the GVF and in some cases better than the GVF.

Digital color reproduction on color television monitors

Color displays used in image processing systems consist of a refresh memory buffer storing digital image data which are converted into analog signals to display an image by driving the primary color channels (red, green, and blue) of a color television monitor. The color cathode ray tube (CRT) of the monitor is unable to reproduce colors exactly due to phosphor limitations, exponential luminance response of the tube to the applied signal, and limitations imposed by the digital-to-analog conversion. In this paper we describe some computer simulation studies (using the U*V*W* color space) carried out to measure these reproduction errors. Further, a procedure to correct for color reproduction error due to the exponential luminance response (gamma) of the picture tube is proposed, using a video-lookup-table and a higher resolution digital-to-analog converter. It is found, on the basis of computer simulation studies, that the proposed gamma correction scheme is effective and robust with respect to variations in the assumed value of the gamma.

Learning Feature Trajectories Using Gabor Filter Bank for Human Activity Segmentation and Recognition

We describe a novel method for human activity segmentation and interpretation in surveillance applications based on Gabor filter-bank features. A complex human activity is modeled as a sequence of elementary human actions like walking, running, jogging, boxing, hand-waving etc. Since human silhouette can be modeled by a set of rectangles, the elementary human actions can be modeled as a sequence of a set of rectangles with different orientations and scales. The activity segmentation is based on Gabor filter-bank features and normalized spectral clustering. The feature trajectories of an action category are learnt from training example videos using dynamic time warping. The combined segmentation and the recognition processes are very efficient as both the algorithms share the same framework and Gabor features computed for the former can be used for the later. We have also proposed a simple shadow detection technique to extract good silhouette which is necessary for good accuracy of an action recognition technique.

Text localization and extraction from complex gray images

We propose two texture-based approaches, one involving Gabor filters and the other employing log-polar wavelets, for separating text from non-text elements in a document image. Both the proposed algorithms compute local energy at some information-rich points, which are marked by Harris' corner detector. The advantage of this approach is that the algorithm calculates the local energy at selected points and not throughout the image, thus saving a lot of computational time. The algorithm has been tested on a large set of scanned text pages and the results have been seen to be better than the results from the existing algorithms. Among the proposed schemes, the Gabor filter based scheme marginally outperforms the wavelet based scheme.

Real-time image denoising algorithm in teleradiology systems - art.no.606304

Denoising of medical images in wavelet domain has potential application in transmission technologies such as teleradiology. This technique becomes all the more attractive when we consider the progressive transmission in a teleradiology system. The transmitted images are corrupted mainly due to noisy channels. In this paper, we present a new real time image denoising scheme based on limited restoration of bit-planes of wavelet coefficients. The proposed scheme exploits the fundamental property of wavelet transform - its ability to analyze the image at different resolution levels and the edge information associated with each sub-band. The desired bit-rate control is achieved by applying the restoration on a limited number of bit-planes subject to the optimal smoothing. The proposed method adapts itself to the preference of the medical expert; a single parameter can be used to balance the preservation of (expert-dependent) relevant details against the degree of noise reduction. The proposed scheme relies on the fact that noise commonly manifests itself as a fine-grained structure in image and wavelet transform allows the restoration strategy to adapt itself according to directional features of edges. The proposed approach shows promising results when compared with unrestored case, in context of error reduction. It also has capability to adapt to situations where noise level in the image varies and with the changing requirements of medical-experts. The applicability of the proposed approach has implications in restoration of medical images in teleradiology systems. The proposed scheme is computationally efficient.

A new FIR filter for image restoration

Image filtering techniques have potential applications in biomedical image processing such as image restoration and image enhancement. The potential of traditional filters largely depends on the apriori knowledge about the type of noise corrupting the image. This makes the standard filters to be application specific. For example, the well-known median filter and its variants can remove the salt-and-pepper (or impulse) noise at low noise levels. Each of these methods has its own advantages and disadvantages. In this paper, we have introduced a new finite impulse response (FIR) filter for image restoration where, the filter undergoes a learning procedure. The filter coefficients are adaptively updated based on correlated Hebbian learning. This algorithm exploits the inter pixel correlation in the form of Hebbian learning and hence performs optimal smoothening of the noisy images. The application of the proposed filter on images corrupted with Gaussian noise, results in restorations which are better in quality compared to those restored by average and Wiener filters. The restored image is found to be visually appealing and artifact-free

Multiple imaging and multichannel optical processing with split lenses

The use of split lenses for multiple imaging and multichannel optical processing is demonstrated. Conditions are obtained for nonoverlapping of multipled images and avoiding crosstalk in the multichannel processing. Almost uniform intensity across the multipled images is an advantage here, while the low ƒ/No. of the split lens segments puts a limit in the resolution in image processing. Experimental results of multiple imaging and of a few multichannel processing are presented.

Believe it or Not! Multicore CPUs can Match GPUs for FLOP-intensive Applications

In this work, we evaluate performance of a real-world image processing application that uses a cross-correlation algorithm to compare a given image with a reference one. The algorithm processes individual images represented as 2-dimensional matrices of single-precision floating-point values using O(n4) operations involving dot-products and additions. We implement this algorithm on a nVidia GTX 285 GPU using CUDA, and also parallelize it for the Intel Xeon (Nehalem) and IBM Power7 processors, using both manual and automatic techniques. Pthreads and OpenMP with SSE and VSX vector intrinsics are used for the manually parallelized version, while a state-of-the-art optimization framework based on the polyhedral model is used for automatic compiler parallelization and optimization. The performance of this algorithm on the nVidia GPU suffers from: (1) a smaller shared memory, (2) unaligned device memory access patterns, (3) expensive atomic operations, and (4) weaker single-thread performance. On commodity multi-core processors, the application dataset is small enough to fit in caches, and when parallelized using a combination of task and short-vector data parallelism (via SSE/VSX) or through fully automatic optimization from the compiler, the application matches or beats the performance of the GPU version. The primary reasons for better multi-core performance include larger and faster caches, higher clock frequency, higher on-chip memory bandwidth, and better compiler optimization and support for parallelization. The best performing versions on the Power7, Nehalem, and GTX 285 run in 1.02s, 1.82s, and 1.75s, respectively. These results conclusively demonstrate that, under certain conditions, it is possible for a FLOP-intensive structured application running on a multi-core processor to match or even beat the performance of an equivalent GPU version.