Artificial retinas — fast, versatile image processors

Kyuma, Kazuo; Lange, E.; Ohta, Jun; Hermanns, Anno; Banish, B.; Oita, Masaya

doi:10.1038/372197a0

Cited by 82 publications

(46 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…To face the challenge of the image processing at video rate, numerous solutions were developed. The electronic retinas appear among one of the most prosperous axes of research [1][2][3][4][5]. Classically, the retinas are characterized by a completely parallel network of analogue operators that receive the signals directly stemming from photosensors.…”

Section: Motivationsmentioning

confidence: 99%

A DSP‐like analogue processing unit for smart image sensors

Dupret

Klein

Nshare

2002

Circuit Theory & Apps

View full text Add to dashboard Cite

SUMMARYAn electronic retina featuring DSP-like programmable analogue processing is addressed. The motivations for designing such an original smart image sensor are accounted for. The architecture of the circuit is described and then the two more important building blocks are detailed. Finally, the practical implementation and tests results are given so as to validate the approach. MOTIVATIONSTo face the challenge of the image processing at video rate, numerous solutions were developed. The electronic retinas appear among one of the most prosperous axes of research [1][2][3][4][5]. Classically, the retinas are characterized by a completely parallel network of analogue operators that receive the signals directly stemming from photosensors. This total parallelism practically authorizes processing within a few milliseconds. Such an approach also o ers very low power consumption along with a high operative density. Besides, the variety of the analogue operators, linear or not, is far richer than those of digital systems. Nevertheless, in real applications, the exploitation of such architectures remains marginal. We attribute this lack of interest to the three following di culties.First of all, for a given silicon area, ÿnding a satisfactory compromise between the resolution and the completeness of the implemented analogue functions is a di cult task. This dilemma can either lead in two impasses. On the one hand, the retinas may o er a rich set of operators but at the expense of the resolution and the ÿll factor. Conversely, the circuit may feature an operative pan too much unadorned to a ord exploitable results.

show abstract

Section: Motivationsmentioning

confidence: 99%

A DSP‐like analogue processing unit for smart image sensors

Dupret

Klein

Nshare

2002

Circuit Theory & Apps

View full text Add to dashboard Cite

show abstract

“…For high speed imaging treatments, the approach of artificial retinas consisting in locating processing at the pixel stage gives some good results [1][2][3]. However, a significant disadvantage lies with this approach: many treatments either remain very difficult to adapt to retina architectures or imply development of specific architectures.…”

Section: Introductionmentioning

confidence: 99%

Reaction–diffusion network for geometric multiscale high speed image processing

Binczak

Sliwa

Jacquir

et al. 2010

Image and Vision Computing

View full text Add to dashboard Cite

a b s t r a c tIn the framework of heavy mid-level processing for high speed imaging, a nonlinear bi-dimensional network is proposed, allowing the implementation of active curve algorithms. Usually this efficient type of algorithm is prohibitive for real-time image processing due to its calculus charge and the inadequate structure for the use of serial or parallel architectures. Another kind of implementation philosophy is proposed here, by considering the active curve generated by a propagation phenomenon inspired from biological modeling. A programmable nonlinear reaction-diffusion system is proposed under front control and technological constraints. Geometric multiscale processing is presented and this opens a discussion about electronic implementation.

show abstract

“…Most of them depend on a camera and a image processor, and are called artificial retinas. 1 Various methods have been used to extract the features of artificial retinas, such as mathematical transforms and neural networks, a'3 These methods really work: they can do edge detection, recognize alphabets, and so on, but they have a common weakness in that the algorithms they use are not parallel, and take a lot of time in computation and training.…”

Section: Introductionmentioning

confidence: 99%

Developing an artificial retina by evolutionary cellular automata and self-organizing neural networks

Zhang

Sugisaka

1999

Artif Life Robotics

View full text Add to dashboard Cite

Retinas are very important for human beings to get information about their environment. In this paper, we propose a new method to build artificial retinas which have many features similar to real ones. We use evolutionary cellular automata to extract some basic characteristics of objects, and use self-organizing neural networks to distinguish different objects. The results indicate a way to get computer vision by artificial life.

show abstract

Artificial retinas — fast, versatile image processors

Cited by 82 publications

References 5 publications

A DSP‐like analogue processing unit for smart image sensors

A DSP‐like analogue processing unit for smart image sensors

Reaction–diffusion network for geometric multiscale high speed image processing

Developing an artificial retina by evolutionary cellular automata and self-organizing neural networks

Contact Info

Product

Resources

About