Design Of Visual Illusion Reproduction System And Its Application In 3D Imaging System

With the improvement of industrial science and technology,the rapid development of display technology has gradually blurred the boundary between image and the real world.By analyzing the three-dimensional information hidden behind the twodimensional image,the biological brain can construct a three-dimensional model from the tissue based on the characteristic information in the image.In order to better highlight the high-dimensional information in the two-dimensional image,the introduction of visual illusion system has a non-negligible impact on the improvement of three-dimensional imaging effect.This research focuses on the key technology involved in the application of visual illusion reproduction system in 3-D imaging system.This paper first introduces the biological mechanism of visual illusion,and the analysis of the receptive field of optic nerve cells.In particular,in view of its biological characteristics of directional selectivity,the nonlinear processing after wavelet decomposition is used for simulation verification.Secondly,based on the fact that the firing of neurons is a dynamic and stable group activity process,the front and rear states and environmental pixels constantly influence and change each other in visual illusion perception,which could be verified and reproduced by the improved Wilson-Cowan neurodynamic formula.Then based on hardware parallelism and pipelining methods,the accelerated IP core was designed for the intensive computation in the direction decomposition and dynamic iteration,and implemented on the PYNQ.Finally,the prototype design of the overall visual illusion reproduction system was completed,and the function was verified in the three-dimensional imaging system.The main work of this research is reflected in:(1)According to the direction selectivity in visual perception,a solution for image multi-directional wavelet component separation is designed.By introducing the principle of trigonometric function transformation,a tightly supported continuous wavelet basis was designed,which is suitable for continuously changing image signals without truncation errors.Three key directional components of visual perception are presented:horizontal,vertical and diagonal.(2)A visual illusion neurodynamic model is designed based on the perception change of visual illusion images,including the design of normalized image processing,the design of Sigmoid activation function based on frequency influence and the design of system stability in the improved Wilson-Cowan visual illusion neurodynamic model.(3)Hardware-accelerated IP cores are designed by hardware parallelism and pipelining methods,including the special IP cores for matrix computation of wavelet transform,the special IP cores for Gaussian convolution of Wilson-Cowan image dynamics model,and the co-invocation design of software and hardware combined with IP cores.This scheme can effectively improve the image processing speed.Finally,the prototype design of the visual illusion reproduction system was built by integrating the visual illusion multi-directional scale processing,visual neurodynamic system and hardware acceleration design,which enhanced the threedimensional brightness information of the scene in the three-dimensional imaging system,and initially verified the feasibility of the whole system scheme.