EMPA Media Technology

Publications 2013

An OCR Concept for Historic Prints

U. Caluori, K. Simon

We present a new concept for the recognition of single characters, the core component of every OCR-System. The recognition is organized as pattern matching with on-the-fly generated patterns corresponding to the glyphs of a given computer font. Thereby, the set of currently considered fonts can be chosen and exchanged arbitrarily which allows a flexible adaption of the software to each specific task, in particular the needs of historic prints.

We presented this paper at Archiving 2013 in Washington, DC where it won the Best Interactive Paper Award.
View Poster (PDF, 8.5M) | View Paper (PDF, 0.5M)

You can find the scanned input images as well as the generated result images here:

Original Scans (320M) | Scans Compressed to JPG (40M)
Results (10M)

Part of OCR Result for Unger Fraktur FontVisualization of results

Interaction improves judgements of gloss

M. Scheller Lichtenauer, P. Schuetz, P. Zolliker

Rendering materials on displays becomes ubiquitous in industrial design, architecture and visualization. Yet, the experience of the material from other modes of perception is missing in that representation. This forces observers to rely on visual cues only while judging material properties. In the present study, we compare judgements of rough and glossy surfaces by interacting and passive observers. We investigate whether observers actively exploring rendered stimuli judge properties differently than observers passively watching renderings. Resulting inter-observer agreement is significantly higher for interacting observers.

Journal of Vision, accepted 10.11.2013

We presented an abstract as a poster at ECVP 2013 in Bremen, Germany: PDF

Sample

Glyph Recognition by Pattern Matching with On-the-fly Generated Patterns

U. Caluori, K. Simon

In this paper we analyze the performance of a new concept for glyph recognition as the core component of an OCR-system. The classification is conducted as pattern matching where the considered patterns are generated on-the-fly with a font rasterization engine. In particular, our approach supports mass-digitizations of historic prints.

We presented this paper at SoftCOM 2013 in Primosten, Split.

You can find the input files, output visualizations and output log files here:

Input Files (15M)
Output Visualizations (16M) | Output Logs (0.1M)


A Cognition-Oriented Approach to Fundamental Frequency Estimation

U. Glavitsch, K. Simon

This paper presents an efficient, two-phase fundamental frequency detection algorithm in the time-domain. In accordance with the human cognitive process it first computes base fundamental frequency estimates, which are verified and corrected in a second step. The verification step proceeds from high-energy stable segments, where reliable estimates are expected, to lower-energy regions. Irregular cases are handled by computing a series of fundamental frequency variants that are evaluated for highest plausibility, in analogy with the hypothesis testing principle of human thinking. As a proof of concept, the algorithm was evaluated on a clean speech database where it shows significantly lower error rates than a comparable reference method.

NLPCS 2013, 10th International Workshop on Natural Language Processing and Cognitive Science, 15 - 16 October, Marseille, France

View full paper (PDF, 335 KB)

F0 Propagation

Similarity Measures for Pattern Matching On-the-fly

U. Caluori, K. Simon

Recently, we presented a new OCR-concept for historic prints. The core part is the glyph recognition based on pattern matching with patterns that are derived from computer font glyphs and are generated on-the-fly. The classification of a sample is organized as a search process for the most similar glyph pattern. In this paper, we investigate several similarity measures which are of vital importance for this concept.

This paper will be presented at ICMV 2013.

You can find the input files, output visualizations and output log files here:

Input Files (15M)
Output Visualizations (15M) | Output Logs (0.1M)

A Similarity MeasureA similarity measure

Detextive: Optical Character Recognition with Pattern Matching On-the-fly

U. Caluori, K. Simon

In this paper we present a new OCR-concept designed for the requirements of historic prints in the context of mass-digitizations. The core part is the glyph recognition based on pattern matching with patterns that are derived from computer font glyphs and are generated on-the-fly. The classification of a sample is organized as a search process for the most similar glyph pattern. This results in consistently good hit rates for arbitrary fonts without any training. In particular, we investigate the performance of our prototype in comparison to popular commercially available OCR-software.

This paper is submitted to Elsevier.

You can find the input and output files here:

Input Files (15M)

Detextive
Output Visualizations (15M) | Output Logs (0.1M)

Abbyy Finereader
Output Files (0.1M)

Adobe Acrobat
Output Files (0.1M)

Google Tesseract
Output Files (0.1M)

Hit Rates for Black Letter Fonts at 600 ppiHit rates for black letter fonts at 600 ppi

Image Quality Assessment using a High Dynamic Range Display

I. Sprow, D. Küpper, Z. Baranczuk, P. Zolliker

In High Dynamic Range imaging (HDR), a remarkable range of compression algorithms has been provided in recent years. With the emergence of new HDR display technology, HDR content can now be displayed directly. We develop a colorimetrically calibrated viewing environment to evaluate the perceived quality of HDR and compressed LDR images. The goal is to use an HDR display instead of the original scenes. Our results give judgment agreement between the original scene and its HDR image version on a HDR display. Furthermore, we investigate whether visual comparisons of simultaneously displayed HDR and LDR content are feasible. We confirm that this set-up is of great use in the development of Tone Mapping Operators on the basis of an applied visual experiment.

12th International AIC Congress - AIC2013
PDF

Psychovisual Test

Creating HDR Video Content for Visual Quality Assessment using Stop-Motion

P. Zolliker, Z. Baranczuk, D. Küpper, I. Sprow,

A complete workflow from capture via processing to the display of video sequences, entirely in high dynamic range (HDR), is described. This workflow is set up to allow for image quality assessment of HDR content with renderings for smaller dynamic ranges. Special emphasis is on the controlled reproduction of lightness and color. Our method allows capturing scenes with a high dynamic range, storing the HDR content in physically meaningful units and displaying it on a commercially available HDR display.

21st European Signal Processing Conference (EUSIPCO-2013)
PDF

Stop-Motion Sequence

High Dynamic Range imaging:
is the game worth the candle?

D. Küpper, Z. Baranczuk, I. Sprow, U. Caluori,
M. Scheller Lichtenauer, K. Simon , P. Zolliker

High dynamic range imaging (HDRi) is a technology concerned with representing a range of luminances larger than state of the art displays and closer to luminance ranges occuring in natural scenes. We investigate whether the approach of evaluating a dynamic range by the number of just noticeable differences contained within does make sense in a HDRi workflow. We found that disturbing effects of neighboring luminances were hardly perceivable on a standard dynamic range display if the background luminance exceeds 5 cd/m².

21st Color Imaging Conference in Albuquerque.
PDF bibtex

Sample