Eric K. Garcia

OCR binarization and image pre-processing for searching historical documents

We consider the problem of document binarization as a pre-processing step for optical character recognition (OCR) for the purpose of keyword search of historical printed documents. A number of promising techniques from the literature for binarization, pre-filtering, and post-binarization denoising were implemented along with newly developed methods for binarization: an error diffusion binarization, a multiresolutional version of Otsus binarization, and denoising by despeckling. The OCR in the ABBYY FineReader 7.1 SDK is used as a black box metric to compare methods. Results for 12 pages from six newspapers of differing quality show that performance varies widely by image, but that the classic Otsu method and Otsu-based methods perform best on average.

Adaptive Local Linear Regression With Application to Printer Color Management

Local learning methods, such as local linear regression and nearest neighbor classifiers, base estimates on nearby training samples, neighbors. Usually, the number of neighbors used in estimation is fixed to be a global ldquooptimalrdquo value, chosen by cross validation. This paper proposes adapting the number of neighbors used for estimation to the local geometry of the data, without need for cross validation. The term enclosing neighborhood is introduced to describe a set of neighbors whose convex hull contains the test point when possible. It is proven that enclosing neighborhoods yield bounded estimation variance under some assumptions. Three such enclosing neighborhood definitions are presented: natural neighbors, natural neighbors inclusive, and enclosing k-NN. The effectiveness of these neighborhood definitions with local linear regression is tested for estimating lookup tables for color management. Significant improvements in error metrics are shown, indicating that enclosing neighborhoods may be a promising adaptive neighborhood definition for other local learning tasks as well, depending on the density of training samples.

Completely Lazy Learning

Local classifiers are sometimes called lazy learners because they do not train a classifier until presented with a test sample. However, such methods are generally not completely lazy because the neighborhood size k (or other locality parameter) is usually chosen by cross validation on the training set, which can require significant preprocessing and risks overfitting. We propose a simple alternative to cross validation of the neighborhood size that requires no preprocessing: instead of committing to one neighborhood size, average the discriminants for multiple neighborhoods. We show that this forms an expected estimated posterior that minimizes the expected Bregman loss with respect to the uncertainty about the neighborhood choice. We analyze this approach for six standard and state-of-the-art local classifiers, including discriminative adaptive metric kNN (DANN), a local support vector machine (SVM-KNN), hyperplane distance nearest neighbor (HKNN), and a new local Bayesian quadratic discriminant analysis (local BDA). The empirical effectiveness of this technique versus cross validation is confirmed with experiments on seven benchmark data sets, showing that similar classification performance can be attained without any training.

Optimized Regression for Efficient Function Evaluation

In many applications of regression, one is concerned with the efficiency of the estimated function in addition to the accuracy of the regression. For efficiency, it is common to represent the estimated function as a rectangular lattice of values—a lookup table (LUT)—that can be linearly interpolated for any needed value. Typically, a LUT is constructed from data with a two-step process that first fits a function to the data, then evaluates that fitted function at the nodes of the lattice. We present an approach, termed lattice regression, that directly optimizes the values of the lattice nodes to minimize the post-interpolation training error. Additionally, we propose a second-order difference regularizer to promote smoothness. We demonstrate the effectiveness of this approach on two image processing tasks that require both accurate regression and efficient function evaluations: inverse device characterization for color management and omnidirectional super-resolution for visual homing.

Gamut Expansion for Video and Image Sets

A semi-automated gamut expansion method is proposed for transforming the colors of video and images to take advantage of extended-gamut displays. In particular, a custom color transformation is learned from an experts enhancement of a single image on an extended gamut display. This methodology allows for the gamut-expansion to be defined in a contextually appropriate way. From the user-enhanced image, we compare defining the gamut expansion by one linear transformation, or by a multi-dimensional LUT which we learn via local linear regression. We show that using the estimated multi-dimensional LUT with tri-linear interpolation (a standard workflow for ICC profiles and color management modules) leads to significantly more pleasant reproduction of skin tones and bright saturated colors.

Color Management of Printers by Regression over Enclosing Neighborhoods

A popular color management standard for controlling color reproduction is the ICC color profile. The core of the ICC profile is a look-up-table which maps a regular grid of device-independent colors to the printer colorspace. To estimate the look-up-table from sample input-output colors, local linear regression has been shown to work better than other methods. An open problem in local linear regression is how to define the locality or neighborhood for each of the local linear regressions. In this paper, new adaptive neighborhood definitions and regularized local linear regression are proposed to address this problem. The adaptive neighborhood definitions enclose the test sample, and are motivated by a result showing they yield bounded estimation variance. An experiment shows that both regularization and the proposed neighborhoods can lead to a significant reduction in error.

Learning custom color transformations with adaptive neighborhoods

Custom color transformations for images or video can be learned from a small set of sample color pairs by estimating a look-up table (LUT) to describe the enhancement and storing the LUT in an International Color Consortium profile, which is a standard tool for color management. Estimating an accurate LUT from a small set of sample color pairs is challenging. Local linear and ridge re- gression are tested on six definitions of neighborhoods for twenty color enhancements and twenty-five color images. Excellent results were obtained with local ridge regression over proposed enclosing neighborhoods, including a variant of Sibsons natural neighbors. The evaluation of the different estimation methods for this task com- pared the fidelity of the learned color enhancement to the original sample color pairs and the presence of objectionable artifacts in enhanced images. These metrics show that enclosing neighbor- hoods are promising adaptive neighborhood definitions for local classification and regression.