[Paper Review] Cognitive image processing: the time is right to recognize that the world does not rest more on turtles and elephants

Emanuel Diamant|arXiv (Cornell University)|Nov 1, 2014

Embodied and Extended Cognition1 references3 citations

TL;DR

This paper argues that traditional human-centered image processing is inadequate for modern Big Data visual workloads, proposing cognitive image processing as a machine-driven paradigm shift. By transitioning from computational data processing to cognitive information processing, the approach enables autonomous, human-like understanding of visual data, resolving ambiguities between data and information through a new cognitive framework for image analysis.

ABSTRACT

Traditional image processing is a field of science and technology developed to facilitate human-centered image management. But today, when huge volumes of visual data inundate our surroundings (due to the explosive growth of image-capturing devices, proliferation of Internet communication means and video sharing services over the World Wide Web), human-centered handling of Big-data flows is impossible anymore. Therefore, it has to be replaced with a machine (computer) supported counterpart. Of course, such an artificial counterpart must be equipped with some cognitive abilities, usually characteristic for a human being. Indeed, in the past decade, a new computer design trend - Cognitive Computer development - is become visible. Cognitive image processing definitely will be one of its main duties. It must be specially mentioned that this trend is a particular case of a much more general movement - the transition from a "computational data-processing paradigm" to a "cognitive information-processing paradigm", which affects today many fields of science, technology, and engineering. This transition is a blessed novelty, but its success is hampered by the lack of a clear delimitation between the notion of data and the notion of information. Elaborating the case of cognitive image processing, the paper intends to clarify these important research issues.

Motivation & Objective

• To address the limitations of human-centered image processing in the face of massive visual data growth.
• To advocate for a paradigm shift from computational data processing to cognitive information processing.
• To clarify the distinction between data and information in visual data processing.
• To establish cognitive image processing as a core function of emerging cognitive computing systems.
• To position cognitive image processing as a foundational component of next-generation AI-driven visual systems.

Proposed method

• Proposes a transition from traditional image processing to a cognitive information-processing paradigm.
• Introduces cognitive computing as a framework for autonomous visual data understanding.
• Emphasizes machine-based processing with human-like cognitive abilities for image analysis.
• Uses the metaphor of 'turtles and elephants' to critique outdated, recursive models of data processing.
• Defines cognitive image processing as a specialized application of broader cognitive computing principles.
• Stresses the need for clear delimitation between data and information to enable effective cognitive processing.

Experimental results

Research questions

• RQ1 How can visual data processing evolve beyond human-centric models to meet Big Data demands?
• RQ2 What distinguishes data from information in the context of image processing?
• RQ3 Why is the current computational data-processing paradigm insufficient for modern visual data workloads?
• RQ4 How can cognitive computing principles be applied to image processing for autonomous understanding?
• RQ5 What role does cognitive image processing play in the broader shift to cognitive information processing?

Key findings

• Traditional image processing is no longer viable for handling the scale of modern visual data due to human processing limitations.
• A paradigm shift from computational data processing to cognitive information processing is both necessary and timely.
• Cognitive image processing is identified as a central function of emerging cognitive computing systems.
• The distinction between data and information remains ambiguous in current systems, impeding progress.
• The paper establishes cognitive image processing as a critical component in the evolution of intelligent visual data systems.
• The transition to cognitive processing enables autonomous, context-aware analysis of visual data at scale.