arXiv subjects

Emanuel Diamant

Publications and source records attributed to Emanuel Diamant.

23 records · Page 2Linked to original sources

arXiv2006

In Quest of Image Semantics: Are We Looking for It Under the Right Lamppost?

In the last years we witness a dramatic growth of research focused on semantic image understanding. Indeed, without understanding image content successful accomplishment of any image-processing task is simply incredible. Up to the recent times, the ultimate need for such understanding has been met by the knowledge that a domain expert or a vision system supervisor have contributed to every image-processing application. The advent of the Internet has drastically changed this situation. Internet sources of visual information are diffused and dispersed over the whole Web, so the duty of information content discovery and evaluation must be relegated now to an image understanding agent (a machine or a computer program) capable to perform image content assessment at a remote image location. Development of Content Based Image Retrieval (CBIR) techniques was a right move in a right direction, launched about ten years ago. Unfortunately, very little progress has been made since then. The reason for this can be seen in a rank of long lasting misconceptions that CBIR designers are continuing to adhere to. I hope, my arguments will help them to change their minds.

cs.CV↗

arXiv2005

Does a Plane Imitate a Bird? Does Computer Vision Have to Follow Biological Paradigms?

We posit a new paradigm for image information processing. For the last 25 years, this task was usually approached in the frame of Treisman's two-stage paradigm [1]. The latter supposes an unsupervised, bottom-up directed process of preliminary information pieces gathering at the lower processing stages and a supervised, top-down directed process of information pieces binding and grouping at the higher stages. It is acknowledged that these sub-processes interact and intervene between them in a tricky and a complicated manner. Notwithstanding the prevalence of this paradigm in biological and computer vision, we nevertheless propose to replace it with a new one, which we would like to designate as a two-part paradigm. In it, information contained in an image is initially extracted in an independent top-down manner by one part of the system, and then it is examined and interpreted by another, separate system part. We argue that the new paradigm seems to be more plausible than its forerunner. We provide evidence from human attention vision studies and insights of Kolmogorov's complexity theory to support our arguments. We also provide some reasons in favor of separate image interpretation issues.

cs.NE↗

arXiv2005

Paving the Way for Image Understanding: A New Kind of Image Decomposition is Desired

In this paper we present an unconventional image segmentation approach which is devised to meet the requirements of image understanding and pattern recognition tasks. Generally image understanding assumes interplay of two sub-processes: image information content discovery and image information content interpretation. Despite of its widespread use, the notion of "image information content" is still ill defined, intuitive, and ambiguous. Most often, it is used in the Shannon's sense, which means information content assessment averaged over the whole signal ensemble. Humans, however,rarely resort to such estimates. They are very effective in decomposing images into their meaningful constituents and focusing attention to the perceptually relevant image parts. We posit that following the latest findings in human attention vision studies and the concepts of Kolmogorov's complexity theory an unorthodox segmentation approach can be proposed that provides effective image decomposition to information preserving image fragments well suited for subsequent image interpretation. We provide some illustrative examples, demonstrating effectiveness of this approach.

cs.CV↗

arXiv2005

Searching for image information content, its discovery, extraction, and representation

Image information content is known to be a complicated and controvercial problem. This paper posits a new image information content definition. Following the theory of Solomonoff-Kolmogorov-Chaitin's complexity, we define image information content as a set of descriptions of imafe data structures. Three levels of such description can be generally distinguished: 1)the global level, where the coarse structure of the entire scene is initially outlined; 2) the intermediate level, where structures of separate, non-overlapping image regions usually associated with individual scene objects are deliniated; and 3) the low-level description, where local image structures observed in a limited and restricted field of view are resolved. A technique for creating such image information content descriptors is developed. Its algorithm is presented and elucidated with some examples, which demonstrate the effectiveness of the proposed approach.

cs.CV↗

arXiv2005

Top-Down Unsupervised Image Segmentation (it sounds like oxymoron, but actually it is not)

Pattern recognition is generally assumed as an interaction of two inversely directed image-processing streams: the bottom-up information details gathering and localization (segmentation) stream, and the top-down information features aggregation, association and interpretation (recognition) stream. Inspired by recent evidence from biological vision research and by the insights of Kolmogorov Complexity theory, we propose a new, just top-down evolving, procedure of initial image segmentation. We claim that traditional top-down cognitive reasoning, which is supposed to guide the segmentation process to its final result, is not at all a part of the image information content evaluation. And that initial image segmentation is certainly an unsupervised process. We present some illustrative examples, which support our claims.

cs.CV↗