Albrecht J. Lindner

Joint statistical analysis of images and keywords with applications in semantic image enhancement

With the advent of social image-sharing communities, millions of images with associated semantic tags are now available online for free and allow us to exploit this abundant data in new ways. We present a fast non-parametric statistical framework designed to analyze a large data corpus of images and semantic tag pairs and find correspondences between image characteristics and semantic concepts. We learn the relevance of different image characteristics for thousands of keywords from one million annotated images. We demonstrate the frameworks effectiveness with three different examples of semantic image enhancement: we adapt the gray-level tone-mapping, emphasize semantically relevant colors, and perform a defocus magnification for an image based on its semantic context. The performance of our algorithms is validated with psychophysical experiments.

Training for Task Specific Keypoint Detection

In this paper, we show that a better performance can be achieved by training a keypoint detector to only find those points that are suitable to the needs of the given task. We demonstrate our approach in an urban environment, where the keypoint detector should focus on stable man-made structures and ignore objects that undergo natural changes such as vegetation and clouds. We use WaldBoost learning with task specific training samples in order to train a keypoint detector with this capability. We show that our aproach generalizes to a broad class of problems where the task is known beforehand.

Measurement of printer MTFs

In this paper we compare three existing methods to measure the Modulation Transfer Function (MTF) of a printing system. Although all three methods use very distinct approaches, the MTF values computed for two of these methods strongly agree, lending credibility to these methods. Additionally, we propose an improvement to one of these two methods, initially proposed by Jang & Allebach. We demonstrate that our proposed modification improves the measurement precision and simplicity of implementation. Finally we discuss the pros and cons of the methods depending on the intended usage of the MTF.

Semantic-Improved Color Imaging Applications: It Is All About Context

Multimedia data with associated semantics is omnipresent in todays social online platforms in the form of keywords, user comments, and so forth. This article presents a statistical framework designed to infer knowledge in the imaging domain from the semantic domain. Note that this is the reverse direction of common computer vision applications. The framework relates keywords to image characteristics using a statistical significance test. It scales to millions of images and hundreds of thousands of keywords. We demonstrate the usefulness of the statistical framework with three color imaging applications: 1) semantic image enhancement: re-render an image in order to adapt it to its semantic context; 2) color naming: find the color triplet for a given color name; and 3) color palettes: find a palette of colors that best represents a given arbitrary semantic context and that satisfies established harmony constraints.

Compensation of printer MTFs

Premilary experiments have shown that the quality of printed images depends on the capacity of the printing system to accurately reproduce details.1 We propose to improve the quality of printed images by compensating for the Modulation Transfer Function (MTF) of the printing system. The MTF of the printing system is measured using the method proposed by Jang and Allebach,2 in which test pages consisting of series of patches with different 1D sinusoidal modulations (modified to improve the accuracy of the results3) are printed, scanned and analyzed. Then the MTF is adaptively compensated in the Fourier domain, depending both on frequency and local mean values. Results of a category judgment experiment show significant improvement as the printed MTF compensated images obtain the best scores.

What impacts skin color in digital photos

Skin colors are important for a broad range of imaging applications to assure quality and naturalness. We discuss the impact of various metadata on skin colors in images, i.e. how does the presence of a metadata attribute influence the expected skin color distribution for a given image. For this purpose we employ a statistical framework to automatically build color models from image datasets crawled from the web. We assess both technical and semantic metadata and show that semantic metadata has a more significant impact. This suggests that semantic metadata holds important cues for processing of skin colors. Further we demonstrate that the refined skin color models from our automatic framework improve the accuracy of skin detection.

Semantic awareness for automatic image interpretation

This thesis aims at developing methods to make digital devices more automatic and intuitive while focusing on image-related applications. We learn associations between image characteristics and keywords with a statistical framework based on large databases of annotated images. Such associations are widely exploitable and we demonstrate two main applications: semantic image enhancement and automatic color naming. Both applications show convincing results and suggest that the framework can be extended to other areas.