|
Information Processing & Management(Impact Factor: 3.892, 27/155)
Summary and Scope:
With the deployment of low-cost sensors, social media platforms, and cloud storage, the tremendous amount of image, video, and textual signals are cheaply available. As a standard tool to analyze these data, quality model has been pervasively used in domains like intelligent systems and 3D rendering. In the past decades, many shallow quality models have been released and commercialized. Despite their success, conventional quality models might be deficiently effective to handle the massive-scale data nowadays.
Potential challenges include (not limited to):
First, owing to the significant progress in deep feature engineering, deep quality models have been proposed and satisfactory performance was received. But deep model is conducted in a black-box manner, how to make it interpretable or transparent to quality modeling, and encoding human subjective wills and perception are still unsolved.
Second, compared to the fully-annotated signals when modeling small-scale data, it is infeasible to label large-scale image/video at pixel-level due to the unaffordable human resources. In practice, only image/video-level labels or partial labels are available. Even worse, sometimes these weak labels are contaminated. Therefore, how to design a noiserobust weakly-supervised learning algorithm to exploring pixel-level quality-related elements is a tough problem.
Third, conventional quality models typically leverage local/global features to evaluate each image/video, where human visual perception cannot be encoded explicitly. Apparently, human visual perception plays a significant role in quality modeling. In the literature, it is difficult to mimic human visual perception, i.e., predicting human gaze behavior and subsequently modeling the visual signal cognition in human brain.
In this special issue, we will focus on the recent progress in image/video/text quality modeling and analytics. We aim to explore interpretable, noise-tolerant, and perceptionaware deep models to enhance quality models. Submissions related to new image/video/text benchmarks for testing the performance of quality models are also welcome. The primary objective for this special issue is to foster focused attention on the latest research progress in this cutting-edge area. We intend to attract researchers and practitioners from both industry and academia.
Topics of interest include (but are not limited to):
o New deep architectures for image/video quality evaluation;
o Deep algorithms for enhancing the shallow-feature-based intelligent systems; o Quality-driven image/video processing techniques;
o New Quality models in management applications;
o Semantic models for deep image/video quality prediction;
o New management tools based on deep quality models;
o New machine learning algorithms for deep media quality modeling;
o Visual quality prediction for photo and video management systems;
o Leveraging human interactions to improve deep quality models;
o Perception-aware quality models for Internet-scale media retrieval;
o Novel deep quality features and their applications in pattern recognition.
o Deep models trained using small samples for quality understanding;
o Novel photo or video retargeting/cropping/re-composition using deep features;
o New datasets, benchmarks, and validation of deep quality models;
o Subjective methodologies to estimate the quality in real-world systems;
o Novel visualization technologies for deep quality features;
Schedule:
• Paper submission due: Dec 1, 2019
• First notification: Feb 1, 2020
• Revision: April 1, 2020
• Final decision: June 15, 2020
• Publication date: Autumn 2020 (Tentative)
|
