Research
Photo Retouching
Automated, Dynamic Assessment of Image Quality Criteria
Motivation
Selecting suitable images for a given image style (also called a briefing) from a large number of source images, for example from a photo shoot, always takes the agency and repro staff involved a great deal of time. Methods based on artificial intelligence are therefore being developed for the automatic assessment of image data according to application-specific image styles. Precisely targeted image data characteristic of a particular style (original and corrected images) form the basis for the development of a neural network that enables dynamic image quality assessment without a formalised (human) description of the respective taste or image style. The assessment works in such a way that arbitrary images are automatically checked, for a given output process, as to whether they correspond to a particular image style and how much effort any necessary retouching (image correction) would require (“traffic light system”).
Scientific Questions
The working hypothesis is that “image-to-image translation” and “image-to-feature translation” are capable of producing such a traffic light system. Since decisions about the suitability of image data are already made manually by experts, sufficient training data is available. This approach is further based on the hypothesis that recognising the relevant features for the gradual match with an image style is sufficient to apply that style successfully to new images, and thus to correct them automatically. This last step is particularly demanding and represents the greatest scientific challenge.
Thesis Topics
New theses in the area of automated, dynamic decisions on image quality criteria are advertised regularly. In addition to the Overview , there are also numerous topics that have not yet been advertised, which will be gladly presented in a personal conversation.
Daten-Upload
To investigate new methods we are building an image database containing both original, unedited image files and edited ones that have been optimised for a particular image style. Such image styles may come from areas such as fashion, tourism or food. In addition, we ask for an assessment of these image files according to quality criteria such as (1 – poorly suited, …, 5 – perfectly suited).
At diesem Link you can upload your images and thus support us in our project.
Partners
- Dr. Andreas Kraushaar, Forschungsinstitut für Medientechnologien e. V. (Fogra), Aschheim near Munich
External Funding
- AiF – Forschungsnetzwerk Mittelstand, “Machine Learning Framework zur dynamischen Entscheidung von Bildqualitätskriterien” Project nr. AiF 21209 N