Hila Nachlieli

Perceptual Segmentation: Combining Image Segmentation With Object Tagging

Human observers understand the content of an image intuitively. Based upon image content, they perform many image-related tasks, such as creating slide shows and photo albums, and organizing their image archives. For example, to select photos for an album, people assess image quality based upon the main objects in the image. They modify colors in an image based upon the color of important objects, such as sky, grass or skin. Serious photographers might modify each object separately. Photo applications, in contrast, use low-level descriptors to guide similar tasks. Typical descriptors, such as color histograms, noise level, JPEG artifacts and overall sharpness, can guide an imaging application and safeguard against blunders. However, there is a gap between the outcome of such operations and the same task performed by a person. We believe that the gap can be bridged by automatically understanding the content of the image. This paper presents algorithms for automatic tagging of perceptual objects in images, including sky, skin, and foliage, which constitutes an important step toward this goal.

Measuring the Quality of Quality Measures

Print quality (PQ) is a composite attribute defined by human perception. As such, the ultimate way to determine and quantify PQ is by human survey. However, repeated surveys are time consuming and often represent a burden on processes that involve repeated evaluations. A desired alternative would be an automatic quality rating tool. Once such quality evaluation measure is proposed, it should be qualified. That is, it should be shown to reflect human assessment. If two of the human opinions conflict, the tool cannot possibly agree with both. Conflicts between human opinions are common, which complicates the evaluation of tools success in reflecting human judgment. There are many optional ways for measuring the agreement between human assessment and tool evaluation, but different methods may have conflicting results. It is, therefore, important to pre-establish the appropriate method for the evaluation of quality-evaluation-tools, a method that takes the disagreement among the survey participants into account. In this paper, we model human quality preference and derive the most appropriate method to qualify quality evaluation tools. We demonstrate the resulting qualification method in a real life scenario-the qualification of the mechanical band meter.

Comprehensive solutions for automatic removal of dust and scratches from images

Dust, scratches, or hair on originals (prints, slides, or negatives) distinctly appear as light or dark artifacts on a scan. These unsightly artifacts have become a major consumer concern. There are several scenarios for removal of dust and scratch artifacts. One scenario is during acquisition, e.g., while scanning photographic media. Another is artifact removal from a digital image in an image editor. For each scenario, a different solution is suitable, with different performance requirements and differing levels of user interaction. This work describes a comprehensive set of algorithms for automatically removing dust and scratches from images. Our algorithms solve a wide range of use scenarios. A dust and scratch removal solution has two steps: a detection step and a reconstruction step. Very good detection of dust and scratches is possible using side information, such as provided by dedicated hardware. Without hardware assistance, dust and scratch removal algorithms generally resort to blurring, thereby losing image detail. We present algorithmic alternatives for dust and scratch detection. In addition, we present reconstruction algorithms that preserve image detail better than previously available alternatives. These algorithms consistently produce visually pleasing images in extensive testing.

Detection of textured areas in natural images using an indicator based on component counts

An algorithm is presented for the detection of textured areas in natural images. Texture detection has potential application to image enhancement, tone correction, defect detection, content classification, and image segmentation. For example, texture detection may be useful for object detection when combined with color models and other descriptors. Sky, e.g., is generally smooth, and foliage is textured. The texture detector presented here is based on the intuition that texture in a natural image is comprised of many components. The measure we develop examines the structure of local regions of the image. This structural approach enables us to detect both structured and unstructured textures at many scales. Furthermore, it distinguishes between edges and texture, and also between texture and noise. Automatic detection results are shown to match human classification of corresponding image areas.

Psychophysical evaluation of banding visibility in the presence of print content

Observing and evaluating print defects represents a major challenge in the area of print quality research. Visual identification and quantification of these print defects becomes a key issue for improving print quality. However, the page content may confound the visual evaluation of print defects in actual printouts. Our research is focused on banding in the presence of print content in the context of commercial printing. In this paper, a psychophysical experiment is described to evaluate the perception of bands in the presence of print content. A number of banding defects are added by way of simulation to a selected set of commercial print contents to form our set of stimuli. The participants in the experiment mark these stimuli based on their observations via a graphical user interface (GUI). Based on the collection of the marked stimuli, we were able to see general consistency among different participants. Moreover, the results showed that the likelihood of an observer perceiving the banding defect in a smooth area is much higher than in a high frequency area. Furthermore, our results also indicate that the luminance of the image may locally affect the visibility of the print defects to some degree.

Masking mediated print defect visibility predictor

Banding is a well-known artifact produced by printing systems. It usually appears as lines perpendicular to the process direction of the print. Therefore, banding is an important print quality issue which has been analyzed and assessed by many researchers. However, little literature has focused on the study of the masking effect of content for this kind of print quality issue. Compared with other image and print quality research, our work is focused on the print quality of typical documents printed on a digital commercial printing press. In this paper, we propose a Masking Mediated Print Defect Visibility Predictor (MMPDVP) to predict the visibility of defects in the presence of customer content. The parameters of the algorithm are trained from ground-truth images that have been marked by subjects. The MMPDVP could help the press operator decide whether the print quality is acceptable for specific customer requirements. Ultimately, this model can be used to optimize the print-shop workflow.

Clustering by Intent: A Semi-Supervised Method to Discover Relevant Clusters Incrementally

Our business users have often been frustrated with clustering results that do not suit their purpose; when trying to discover clusters of product complaints, the algorithm may return clusters of product models instead. The fundamental issue is that complex text data can be clustered in many different ways, and, really, it is optimistic to expect relevant clusters from an unsupervised process, even with parameter tinkering. We studied this problem in an interactive context and developed an effective solution that re-casts the problem formulation, radically different from traditional or semi-supervised clustering. Given training labels of some known classes, our method incrementally proposes complementary clusters. In tests on various business datasets, we consistently get relevant results and at interactive time scales. This paper describes the method and demonstrates its superior ability using publicly available datasets. For automated evaluation, we devised a unique cluster evaluation framework to match the business users utility.

Color-dependent banding characterization and simulation on natural document images

Print defects like banding from a digital press involve not only luminance variation, but also chrominance variation. As digital presses place one color separation at a time, the contrast and spatial pattern of the print defects are color-space dependent. Characterizing the color-dependent features of the banding signal enables us to simulate the banding on natural document images in a more accurate way that matches the characteristics of the banding generation mechanism within the digital press. A framework is described for color-dependent banding characterization including the following steps: printing and scanning uniform patches that sample colorant combinations throughout the input document sRGB color space, extracting banding signals in the CMYK color space of the target device, and modeling the banding features in a perceptually uniform color space. We obtain a full banding features LUT for every color point in the input sRGB space by interpolating banding features extracted from measured color points. The color-dependent banding simulation framework is developed based on the banding features LUT. Using the information contained in this LUT, a single banding prototype signal is modulated in a color-space-dependent fashion that varies spatially across the natural document image. Proper execution of the framework of banding characterization and simulation requires careful calibration of each system component, as well as implementation of a complete color management pipeline.