May 9th, 2012

"Applications of Interactive Computer Vision" by Junhong GAO

We present two applications of interactive computer vision. The first involves an efficient method for producing picture legends for group photos. This approach combines face detection with human shape priors into an interactive selection framework to allow users to quickly segment the individuals in a group photo. Results obtained by our method are better than those obtained by general selection tools and can be produced in a fraction of the time. Our second method is a tool for correcting errors in panoramic images. In particular, we describe two features:

  1. a seam-editing tool that allows the user to modify blending seams in a local manner
  2. a content-aware snapping tool to help the user better align local image content between overlapping images 

We demonstrate the effectiveness of our tool on several examples that are tedious to achieve using existing photo-editing softwares.


"A New In-Camera Imaging Model for Color Computer Vision and its Application" by Haiting LIN

We present a study of the in-camera image processing through an extensive analysis of more than 10,000 images from over 30 cameras. The goal of this work is to investigate if image values can be transformed to physically meaningful values, and if so, when and how this can be done. From our analysis, we found a major limitation of the imaging model employed in conventional radiometric calibration methods and propose a new in-camera imaging model that fits well with today’s cameras. With the new model, we present associated calibration procedures that allow us to convert sRGB images back to their original CCD RAW responses in a manner that is significantly more accurate than any existing methods. Additionally, we show how this new imaging model can be used to build an image correction application that converts an sRGB input image captured with the wrong camera settings to an sRGB output image that would have been recorded under the correct settings of a specific camera.
We also describe a method to construct a sparse lookup table (LUT) that is effective in modeling the camera imaging pipeline that maps a RAW camera image to its sRGB output based on the new aforementioned color processing model. We show how to construct a LUT using a novel nonuniform lattice regression method that adapts the LUT lattice to better fit the underlying 3D function which was previously formulated as a RBF function. Our method offers not only a performance speedup of an order of magnitude faster than RBF, but also a compact mechanism to describe the imaging pipeline.


"Marked Point Process for Automatic Neuronal Network Extraction" by Sreetama BASU

Neuronal structures are intricately related to their functions. Study of the neuronal structures reveals healthy and pathologic conditions, crucial to understanding how the Brain works. Current advances in microscopy techniques produce huge volume of data where manual reconstruction and analysis may take several years. Moreover, most of this data is sparse; hence digital reconstructions capturing the essential structural information of the neuronal networks provide ease of archiving, exchanging and analysing. The lack of powerful computational tools to automatically reconstruct neuronal arbors has emerged as a major technical bottleneck in neuroscience research. This work extends the Marked Point Process methodology, which has been proved to be an efficient framework for network extraction in 2D, to 3D neuronal network extraction from microscopy image stacks. The optimization process considers a multiple birth and death dynamics embedded in a simulated annealing scheme. To speed up the convergence a birth map based on the projection of the neuronal processes is considered.


"Cognitive Vision: Towards Ontology-driven Medical Image Processing" by Olivier MORERE

In close collaboration with AGFA Healthcare and La Pitié Salpêtrière Hospital, Paris, France, IPAL’s MICO (COgnitive virtual MIcroscopy) platform aims at developing a cognition-driven visual explorer for histopathology, particularly for breast cancer grading, supported by dynamic semantic annotation and medical ontology. 
The analysis capabilities and results are made available to the pathologist through a platform combining virtual microscopy and cognitive reasoning. This allows the medical staff to interact with the platform at the appropriate level of abstraction. The platform should combine multi-modal histopathological images, multi-scale whole slide image (WSI) exploration & analysis, and medical knowledge representation & inference using ontologies. 
Semantic tools should be used to drive image exploration & analysis. A semantic profile should be provided to each algorithm, allowing high flexibility and good knowledge gathering. Medical knowledge should also be integrated into MICO, improving it’s abilities to interact with the histopathologist users, helping them to make the right choices.
How Semantic Web technologies could be useful in frameworks having numerous and dynamic data sources, interacting with computer science inexperienced users in order to provide them proven answers to their problems ?