Michael A. Pratt

Multiresolution superpixels for visual saliency detection

Salient regions are those that stand out from others in an image. We present an algorithm to detect salient regions in an image that is represented as superpixels at a number of resolutions. Superpixels are segments generated by oversegmenting an image and they form a perceptually meaningful representation that preserves the underlying image structure. The novelty of our method is the ranking of a superpixel by its dissimilarities with respect to other superpixels and highlighting the statistically salient region proportional to their rank. This is based on the premise that salient region group together and they stand out. We tested our method using standard data sets containing images of varied complexity and compared the results to ground truth data. Our results show that our saliency detection algorithm is robust to changes in color, object size, object location in image and background type.

Saliency Detection using Geometric Context Contrast Inferred from Natural Images

Image saliency detection using region contrast is often based on the premise that salient region has a contrast with the background which becomes a limiting factor if the color of the salient object background is similar. To overcome this problem associated with single image analysis, we propose to collect background regions from a collection of images where generative property of, say, natural images ensures that all the images are spun out of it hence negating any bias. Background regions are differentiated based on their geometric context where we use the ground and sky context as background. Finally, the aggregated map is generated using color contrast between the superpixels segments of the image and collection of background

Texture‐based image steganalysis by artificial neural networks

Purpose – The purpose of this paper is to present research results in analyzing image contents to improve the accuracy of using an artificial neural network (ANN) to detect embedded data in a digital image.Design/methodology/approach – A texture measure based on the MPEG‐7 texture descriptor is applied to assess the local texture amount. Those image blocks with high texture are masked out and the remaining blocks with low texture are used to derive features for an ANN to classify an image as embedded or clear. The high‐texture blocks are not discarded and can be tested independently for embedded data.Findings – By masking out the high‐texture image blocks, an ANN has improved detection performance especially when the original embedding rate is low. Bypassing the low‐texture image blocks do not pay off for a steganographer because the effective embedding rate in the high‐texture blocks is driven higher.Research limitations/implications – Hidden data detectors should take the image content into account in o...

Visual saliency approach to anomaly detection in an image ensemble

Visual saliency is a bottom-up process that identifies those regions in an image that stand out from their surroundings. We oversegment an image as a collection of “super pixels” (SPs). Each SP is salient if it is different in color from all other SPs and if its most similar SPs are nearby. We test our method on image sequences collected by a vehicle. We consider an SP in a frame as salient if it stands out from all frames in a collection that consists of an ensemble of images from different road segments and a sequence of immediate past frames.

Predicting Hospital Safety Measures using Patient Experience of Care Responses.

To make healthcare more cost effective, the current trend in the U.S. is towards a hospital value-based purchasing program. In this program, a hospital’s performance is measured in the safety, patient experience of care, clinical care, and efficiency and cost reduction domains. We investigate the efficacy of predicting the safety measures using the patient experience of care measures. We compare four classifiers in the prediction tasks and concluded that random forest and support vector machine provided the best performance.

Analysis of geographical variations of healthcare providers performance using the empirical mode decomposition

Performance of healthcare providers such as hospitals varies from one locale to another. Our goal is to study whether there is a geographical pattern of performance using metrics reported from over 3,000 hospitals distributed across the U.S. Empirical mode decomposition (EMD) is an effective analysis tool for nonlinear and non-stationary signals. It decomposes a data sequence into a series of intrinsic mode functions (IMFs) along with a residue sequence that represents the trend. Each IMF has zero local mean and has exactly one zero crossing between any two consecutive local extrema. An IMF can be used to assess the instantaneous frequency. Reconstruction of a signal using the residue and those IMFs of the lower frequency can reveal the underlying pattern of the signal without undue influence of the higher frequency fluctuations of the data. We used a space-filling curve to turn a set of performance metrics distributed irregularly across the two-dimensional planar surface into a one-dimensional sequence. The EMD decomposed a set of hospital emergency department median waiting times into 9 IMFs along with a residue. We used the residue and the lower frequency IMFs to reconstruct a sequence with fewer fluctuations. The sequence was transformed back to a two-dimensional map to reveal the geographical variations.

Visually salient features for highway scene analysis

In applications involving autonomous vehicle or camera assisted driving it is important to have a generic prior understanding of the highway scene. We present a visually salient feature-based approach for road image understanding. We use salient features for object localization, near-far distinction and urban-rural scene classification; these tasks have such applications as in adjusting a vehicles speed. We empirically verify the efficacy and assess the performance of each task.

Earth mover's distances of feature vectors in large data analyses

The earth movers distance (EMD) measures the difference of two feature vectors that is related to the Wasserstein metric defined for probability distribution functions on a given metric space. The EMD of two vectors is based on the cost of moving the content of individual elements of an anchor vector to match the distribution of a target vector. The EMD is a solution to a transportation problem. We present results of using EMD in large data analysis problems such as those for health data and image data.

Compressive sampling approach to visual attention in image scene analysis

Many image scene analysis applications require a computational approach to visual attention. The foreground in these applications is typically sparse in spatial support. Compressive sampling enables an approach to reconstruct the sparse map of image regions that stand out from the background using fewer measurements. A convex optimization algorithm, for instance, can be used to recover the sparse map in the wavelet domain. Besides being sparse in the transform domain, the background of natural images has an interesting property that the amplitude of the averaged Fourier spectrum is approximately proportional to the inverse of the frequency. This further enables us to approximate an average background signal for extracting the out-of-ordinary foreground signal corresponding to objects of interest.