We propose a novel approach to learn and recognize natural scene categories. Unlike previous work, it does not require experts to annotate the training set. We represent the image of a scene by a collection of local regions, denoted as codewords obtained by unsupervised learning. Each region is represented as part of a "theme". In previous work, such themes were learnt from hand-annotations of experts, while our method learns the theme distributions as well as the codewords distribution over the themes without supervision. We report satisfactory categorization performances on a large set of 13 categories of complex scenes.

/pdf/a-bayesian-hierarchical-model-for-learning-natural-scene-zzqtm1es9k.pdf

A Bayesian hierarchical model for learning natural scene categories

This paper deals with the single image scale-up problem using sparse-representation modeling. The goal is to recover an original image from its blurred and down-scaled noisy version. Since this problem is highly ill-posed, a prior is needed in order to regularize it. The literature offers various ways to address this problem, ranging from simple linear space-invariant interpolation schemes (e.g., bicubic interpolation), to spatially-adaptive and non-linear filters of various sorts. We embark from a recently-proposed successful algorithm by Yang et. al. [1,2], and similarly assume a local Sparse-Land model on image patches, serving as regularization. Several important modifications to the above-mentioned solution are introduced, and are shown to lead to improved results. These modifications include a major simplification of the overall process both in terms of the computational complexity and the algorithm architecture, using a different training approach for the dictionary-pair, and introducing the ability to operate without a training-set by boot-strapping the scale-up task from the given low-resolution image. We demonstrate the results on true images, showing both visual and PSNR improvements.

On single image scale-up using sparse-representations

Alzheimer's disease (AD) and antecedent factors associated with AD were explored using amyloid imaging and unbiased measures of longitudinal atrophy in combination with reanalysis of previous metabolic and functional studies. In total, data from 764 participants were compared across five in vivo imaging methods. Convergence of effects was seen in posterior cortical regions, including posterior cingulate, retrosplenial, and lateral parietal cortex. These regions were active in default states in young adults and also showed amyloid deposition in older adults with AD. At early stages of AD progression, prominent atrophy and metabolic abnormalities emerged in these posterior cortical regions; atrophy in medial temporal regions was also observed. Event-related functional magnetic resonance imaging studies further revealed that these cortical regions are active during successful memory retrieval in young adults. One possibility is that lifetime cerebral metabolism associated with regionally specific default activity predisposes cortical regions to AD-related changes, including amyloid deposition, metabolic disruption, and atrophy. These cortical regions may be part of a network with the medial temporal lobe whose disruption contributes to memory impairment.

https://www.jneurosci.org/content/jneuro/25/34/7709.full.pdf

Molecular, Structural, and Functional Characterization of Alzheimer's Disease: Evidence for a Relationship between Default Activity, Amyloid, and Memory

The linear time-discrete state-space model is generalized from single-dimensional time to two-dimensional space. The generalization includes extending certain basic known concepts from one to two dimensions. These concepts include the general response formula, state-transition matrix, Cayley-Hamilton theorem, observability, and controllability.

A discrete state-space model for linear image processing

In this paper, we propose a new method for indexing large amounts of point and spatial data in high-dimensional space. An analysis shows that index structures such as the R*-tree are not adequate for indexing high-dimensional data sets. The major problem of R-tree-based index structures is the overlap of the bounding boxes in the directory, which increases with growing dimension. To avoid this problem, we introduce a new organization of the directory which uses a split algorithm minimizing overlap and additionally utilizes the concept of supernodes. The basic idea of overlap-minimizing split and supernodes is to keep the directory as hierarchical as possible, and at the same time to avoid splits in the directory that would result in high overlap. Our experiments show that for high-dimensional data, the X-tree outperforms the well-known R*-tree and the TV-tree by up to two orders of magnitude.

/pdf/the-x-tree-an-index-structure-for-high-dimensional-data-2fnnrt8qnu.pdf

The X-tree: an index structure for high-dimensional data

Image processing techniques find applications in many areas, chief among which are image enhancement, pattern recognition, and efficient picture coding. Some aspects of image processing are discussed--specifically: the mathematical operations one is likely to encounter, and ways of implementing them by optics and on digital computers; image description; and image quality evaluation. Many old results are reviewed, some new ones presented, and several open questions are posed.

O.J. Tretiak

Papers

Image processing