About this Research Topic
The field of image analysis, ranging from segmentation to filtering and even compression, has proposed a lot of methodologies for half a century. A recurrent question has been posed about the evaluation and the comparison of old and recent techniques, to have a clever understanding of their applicability and a precise assessment of their robustness in various contexts: medical imaging and computer aided diagnostic, intelligent video surveillance, automatic animal behavior analysis, motion capture, and many more. Our Frontiers Research Topic ‘SPECIAL’ aims at gathering and largely communicating by open access recent works about the evaluation and comparison of image analysis algorithms, and the proposition of new benchmarks and datasets to help research communities to improve their concepts and developments. Another related problem addressed in SPECIAL is the automatic or interactive generation and annotation of data and its associated ground truth, since collecting this information by hand might be a very tedious and time consuming task.
Several benchmarks already exist (for example, PASCAL-VOC for computer vision tasks, SLIVER for liver segmentation, HMDB for Human action recognition, etc.), and challenges are organized in international conferences to help communities in solving those problems. ‘SPECIAL’ should be the opportunity for researchers either to propose methodologies that significantly overtake the others by considering a wide range of approaches from state-of-the-art, or to introduce new benchmarks, data-sets, automatic or interactive data and ground truth generation techniques, all offering new challenges for future contributors.
Our main application concerns medical contexts, where the annotation of image data may reveal differences between experts (radiologists for example), and thus make automatic evaluation more complex. Moreover, in these contexts, images and so graphical objects are represented through multiple modalities and dimensions (namely 2D, 2D+t, 3D, 3D+t). Besides these contexts, the submissions to SPECIAL can consider any other application.
Topics of interest of our Frontiers Research Topic ‘SPECIAL’ include, without being exhaustive:
- Evaluation and comparison of algorithms, quality assessment, robustness
- Automatic or interactive data generation and annotation of ground truth
- Benchmarking: datasets, similarity metrics, diffusion (websites, challenges …), evaluation softwares, etc.
- Medical imaging, computer aided diagnostic, treatment, surgery, etc.
- Other applications: video-surveillance, nature, etc.
Keywords: Image processing, computer cision, medical imaging, evaluation, benchmarking
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.