The image-comparison workstation is a versatile tool for comparative assessment of image quality. At x2 magnification, images compressed with either JPEG or WTCQ algorithms were indistinguishable from unaltered original images for most observers at compression ratios between 8:1 and 16:1, indicating that 10:1 compression is acceptable for primary image interpretation.

Gaussian pyramid, Laplacian pyramid, and subband An object recognition approach based on concurrent pyramid [2]. coarse-and-fine matching using a multi-layer HopfieldThe most immediate utility of a multi-resolution neural network is presented. The The proposed coarse-and-fine strategy is implemented by utilizing a multi-layer Hopfield neural net-1 Introduction work. The single layer Hopfield neural network from which it derives has been used in a wide range of apObject recognition has emerged as a subject of wide plications, such as vision tasks. Vision tasks can be research interest during the last decade'. Two comformulated as an optimization problem where an en-0 mon themes characterizing much of the recent work ergy function is minimized. The search for its global , if have been the use of a priori information in the form minimum can be implemented through a Hopfield neuof models and constraints, and the incorporation of the ral network with interconnection weights generating A-d most current image processing tools to enhance recogan equivalent energy function. Unfortunately, there 1 nition performance. In this spirit, the objective of the typically exist multiple local mininma in such energy 3.present. study is to explore the use of nu lti-resolution functions due to its non-convexity and its argument 0 (pyramidal) image representation in the context of rehigh dimensionality, and a gradient descent procedure cently reported neural network implementation techis vulnerable to early termination. The ilopfield net-4 !I nology, with the goal of faster and more robust, autowork can get trapped in any of these local minima Smated object recognitio, performance. depending on the initial states of the network and the It is natural to seek object recognition cues conway it selects the sequence-by which the states of the A : currently at several resolution levels. Multi-resolution neurons are updated.• el image representation and processing is a well known In this paper, a concurrent (coarse-and-fine) multiimage analysis methodology. A multi-resolution imresolution model-based object recognition technique is age representation can be viewed as an image pyraproposed using a multi-layer Hopfield neural network mid. Important classes of image pyramids include the to alleviate some of these problems which arise when a single layer Hopfield network is utilized.

A multi-institution effort was conducted to assess the visual quality performance of various JPEG 2000 (Joint Photographic Experts Group) lossy compression options for medical imagery. The purpose of this effort was to provide clinical data to DICOM (Digital Imaging and Communications in Medicine) WG IV to support recommendations to the JPEG 2000 committee regarding the definition of the base standard. A variety of projection radiographic, cross sectional, and visible light images were compressed-reconstructed using various JPEG 2000 options and with the current JPEG standard. The options that were assessed included integer and floating point transforms, scalar and vector quantization, and the use of visual weighting. Experts from various institutions used a sensitive rank order methodology to evaluate the images. The proposed JPEG 2000 scheme appears to offer similar or improved image quality performance relative to the current JPEG standard for compression of medical images, yet has additional features useful for medical applications, indicating that it should be included as an additional standard transfer syntax in DICOM.

In low light conditions, visible light face identification is infeasible due to the lack of illumination. For nighttime surveillance, thermal imaging is commonly used because of the intrinsic emissivity of thermal radiation from the human body. However, matching thermal images of faces acquired at nighttime to the predominantly visible light face imagery in existing government databases and watch lists is a challenging task. The difficulty arises from the significant difference between the face's thermal signature and its visible signature (i.e. the modality gap). To match the thermal face to the visible face acquired by the two different modalities, we applied face recognition algorithms that reduce the modality gap in each step of face identification, from low-level analysis to machine learning techniques. Specifically, partial least squares-discriminant analysis (PLS-DA) based approaches were used to correlate the thermal face signatures to the visible face signatures, yielding a thermal-to-visible face identification rate of 49.9%. While this work makes progress for thermal-to-visible face recognition, more efforts need to be devoted to solving this difficult task. Successful development of a thermal-to-visible face recognition system would significantly enhance the Nation's nighttime surveillance capabilities.

Recovering the shape and reflectance of non-Lambertian surfaces remains a challenging problem in computer vision since the view-dependent appearance invalidates traditional photo-consistency constraint. In this paper, we introduce a novel concentric multi-spectral light field (CMSLF) design that is able to recover the shape and reflectance of surfaces with arbitrary material in one shot. Our CMSLF system consists of an array of cameras arranged on concentric circles where each ring captures a specific spectrum. Coupled with a multi-spectral ring light, we are able to sample viewpoint and lighting variations in a single shot via spectral multiplexing. We further show that such concentric camera/light setting results in a unique pattern of specular changes across views that enables robust depth estimation. We formulate a physical-based reflectance model on CM-SLF to estimate depth and multi-spectral reflectance map without imposing any surface prior. Extensive synthetic and real experiments show that our method outperforms stateof-the-art light field-based techniques, especially in non-Lambertian scenes.