PhD Position at the Center for Mathematical Morphology - Text Analysis From Receipts Images and Videos
(CIFRE agreement with SKEROU http://skerou.com)
Description: Text recognition from mobile phone is of quickly growing importance. Differences abound between trying to
find and read a limited quantity of text on a frame that may include anything, and reading a large amount of text on a frame that is known
to contain mainly text. Skerou focuses on the latter, and aims at analyzing text from grocery receipts.
Skerou is a start-up whose goal is to empower consumers handing them the data about their consumption. With a simple picture (or video) of
their receipt, Skerou finds all the items purchased reading, but also understanding, the text of the receipt, and gives that information
back to the consumer. The Skerou application then helps consumers make better decisions based on their habits, following their spending,
automating their shopping list reminding them what they should soon renew, or letting them use promotions without any couponing.
The goal of this PhD is to develop the whole image processing chain, from the receipt image or video to its semantic interpretation. The
following topics, among others, will be addressed:
- Image correction of specific mobile biases: geometric deformations, extended depth of focus, illumination correction;
- Adaptive filtering exploiting receipt characteristics in order to optimize OCR processing;
- Video processing of very long receipts, simplifying the data acquisition process for the end user;
- Super-resolution techniques in order to enhance image quality.
Place: Center for mathematical morphology. Mines ParisTech. Fontainebleau. (http://cmm.ensmp.fr)
- Good knowledge of image processing and machine learning techniques;
- Good programming skills (C++);
- Fluent English;
Application file: a detailed résumé, a list of studies, a transcript of your academic records, a motivation letter
and a list of persons to be contacted for recommendation letters. For further details, please contact Gregory Thurin (firstname.lastname@example.org)
or Beatriz Marcotegui (Beatriz.Marcotegui@mines-paristech.fr)
Research field: Morphological image segmentation, multi-camera image processing.
Place: Centre de Morphologie Mathématique (CMM), Mines Paristech, Fontainebleau.
Duration: 18 months, for immediate appointment
Context: This offer is part of two industrial vision projects. The first one, which is in its last part,
is headed by VI Technology, a company specialized in electronic cards automatic control. The CMM part consists in defining and
developing image analysis and segmentation tools in order to improve and speed up the control process. The second project is an
ENIAC one (http://www.eniac.eu), gathering about 20 partners among which Philips Healthcare, Grass Valley and Bosch. The CMM part
consists in defining low level analysis tools, which will be integrated in multi-view scene presentation.
Subject: The automatic control of electronic cards aims at highlighting various defects that may occur during the
manufacturing process. This control is carried out in 2 steps. The first one consists in programming the control machine to point out
the defect types to be searched for and where. Some tools have already been developed by the CMM, they will have to be complemented
and finalized. The second step consists in the control itself. Various image processing tools will have to be elaborated in order
to improve the control efficiency and speed.
The second project is of an exploratory nature and consists in adding some intelligence in the acquisition system sensors,
in various fields such as broadcasting, medical radiography and video-control. The CMM will have to provide real-time segmentation
tools to be applied to large scale images. These tools will be used by our partners to offer new services on the applicative platform.
Required skills: We are looking for a motivated and dynamic applicant, with a very deep knowledge of image processing,
more particularly in mathematical morphology, with a good command of image processing tools. Moreover, the person will know various
programming languages (Python, C/C++ for Linux and Windows) in order to master the porting of the algorithms developed in the framework
of this project into our partners' environment. A previous experience with GPU or FPGA based image processing accelerators would be
an asset. The applicant must be aware that this applied research subject demands at the same time a great deal of dynamism, high capacities
for adaptation and team work. A very good knowledge of English is compulsory as it is the language used with our partners. A previous
experience in a European or ANR project would also be an advantage.
Please e-mail your motivation letter to: Serge.Beucher@mines-paristech.fr,