I.V. Zhigulina – Ph. D. (Eng.), Associate Professor, Professor, Department of Mathematics, MESC «Zhukovsky–Gagarin Air Force Academy» (Voronezh)
This paper proposes an approach to the pre-processing of images and video sequences, which allows to implement the mechanisms inherent in the retina. Namely, important features of the retina, as the on-off centers and receptive fields of ganglion cells, are consi-dered. Receptive fields better react to moving objects and the edges of objects. Such preprocessing in the retina reduces several orders of information flow coming further into the visual cortex of the brain. It promotes effective analysis and understanding of the real scenes images in the brain.
Each receptive field of the ganglion cells is divided into two parts: the center and periphery. They react to light in the opposite way. All image\'s field also divide into areas of similar configuration. Each of these areas is represented as a circle or a square concentric with the central pixel. The processing can be carried out along any line passing through the central pixel structure. Therefore, in this paper we consider the one-dimensional bizonal structure.
The movement identification can be achieved by the use of the energy characteristics (interframe increment harmonics amplitudes of the energy spectrum). Significant improvement in detection performance is observed after zeroing certain pixels in the image. This is due to the appearance of additional summands in the equations for the energy characteristics. The summands are associated with zeroed pixels. These pixels may belong to different parts of the image (bright or dark). This fact determines of the summands values. Zeroing of video signal counts is a «switching» of corresponding pixels (ON- zone formation), all other pixels can be considered as «not enabled» (OFF-zone formation).
Special functions are defined for bizonal one-dimensional structure. They describe the reaction of the central and peripheral zone to the input action. Such structure can be considered as the receptive field, because it has two zones, and each zone responds to part of the image, which getting on it. It is possible to make ON-processing in each zone (to nullify group of pixels) and to receive ON-OFF zone (center is «enabled») or OFF-ON zone (periphery is «enabled»). The responses of each zone can be analyzed both separately and together. Analysis of the bizonal structure reaction allows identifying the static and dynamic objects.
Bradley C., Abrams J., Geisler W.S. Retina-V1 model of delectability
across the visual field // J. Vis. 2014. № 14(22). P. 1−22.
Tamietto M. et. al. Unseen Facial and Bodily Expressions Trigger
Fast Emotional Reactions // Proceedings of the National Academy of Sciences USA.
20 October 2009. V. 106. № 42. P. 17661−17666.
KHjubel D. Glaz, mozg, zrenie: Per. s angl. M.: Mir. 1990.
Fiziologija cheloveka ( Human Physiology): Per. s angl. Izd.
2-e, pererab. i dop. V 3-kh tomakh. CHast III. Obshhaja i specialnaja sensornaja
fiziologija / Pod red. R. SHmidta i G. Tevsa. M.: Mir. 1996. T. 1.
Bogoslovsky A., Zhigulina I. A way of energy analysis for image
and video sequence processing // Computer vision in control systems‑1. ISRL. / Eds.
Favorskaya M.N., Jain L.C. V. 73. P. 183−210. Switzerland: Springer.
Bogoslovsky A., Zhigulina I., Maslov I., Mordovina T.: Frequency
Characteristics for Video Sequences Processing // Smart Innovation, Systems and
Technologies. V. 40: Intelligent Interactive Multimedia Systems and Services.
P. 149−160. Switzerland: Springer. 2015.