The Noise Simulation Of Camera Imaging In Virtual Environment

Camera simulation is one of the important technologies in the field of image processing. It is expected to solve many problems in the actual production and life. For example, by the simulation of the car camera d uring running, people can improve the internal software of vehicles to make vehicles travel safer. Currently, camera simulation is an indispensable part of virtualization projects in companies and universities. During the process of camera image simulation, the noise may cause the distortion of images, causing an inaccurate detection and thus many security risks. Therefore, the noise simulation is necessary for the practical applications for the study of driverless cars and intelligent noise reduction experiments.Herein, inspired by several previous models of noise simulation, we propose a method to simulate the camera image noise in virtual environment based on the principle of how the sensor arrays of camera capture images. The simulation includes the following two aspects: the thermal noise of the sensors as internal factors and the noise from photon impact and spectral sensitivity as external factors. For the thermal noise of the sensors, which contains the average noise level of camera noise, time-related noise fluctuation and spatial-related noise fluctuation, the random fluctuation values was added onto the camera average noise to fit the total thermal noise. Then we made a comprehensive analysis of the thermal noise to describe the influences of the noise fluctuation on the three channels(red, green and blue) on the time dimension and spatial dimension, respectively. For the noise from spectral sensitivity, we investigated the different effects of photon impact on the different channels under a certain intensity of light. By this way, more detailed tuning or correction could be achieved, which is promising for the simulation of the noise under the real condition.In order to verify the accuracy of this noise simulation method, the noise experiments of real images were carried out and compared with the simulation results by the edge detection and Hough transform detection circle. This paper first sets the sensor thermal noise experiment in the dark environment, and tests the sensor thermal noise level, time dimension and spatial dimension variance. Then in the normal light condition, using the test camera needs to set the light parameters. At the moment, it needs to test the camera spectral sensitivity. Finally, this paper offers the method to fit integrated noise that come from the all noise. Adding the integrated noise to the lossless image in the virtual environment completes perfectly.