With the development of the information technique, more and more optical cameras are utilized by the department of national defense or the department of industry. The optical camera plays a more and more important role in the fields of surveillance, vehicle navigation and firepower control. The advantages of the optical camera include its stability of data capture, the low cost of the system implementation, etc. However, the imaging definition of the optical camera is affected by the environment factors easily. As a result the image quality of the optical camera cannot be guaranteed to be used broadly. The image with good quality is not only benefit to view by people but also good for the subsequent processing of the image or computer vision algorithms. In this dissertation we hold the concept that for an outdoor-worked optical camera, it is necessary to implement the image quality evaluation primarily before any processings of different algorithm. That is to say, for the sequential image, if we evaluate the image quality for each frame and connect all the evaluation result together in the time axis, we can mine the change law of the imaging definition by selecting the proper analysis technique of time series; and it will surely give some guidance for the subsequent computation. The main contributions of this dissertation include: 1) For the problem of image quality evaluation, we improve and design some blind evaluation metrics of the image quality evaluation, and we also present a novel interactive image quality evaluation method. On one hand, our method emphasizes that the design of the objective metric should be given a feedback by the result of the subjective evaluation. So we use the canonical correlation analysis method to build the connection between the results of the subjective and those of the objective evaluation. On the other hand, our method also claim the evaluation of the image quality should consider the special visual task. So we utilize the apparatus of eye tracking system to search the region of the attention and add the attention mechanism into the model of the objective image quality metric. 2) For the problem of the stochastic modeling, we researches the basic modeling technique of the classical Autoregressive Moving Average (ARMA) model, the Autoregressive Conditional Heteroscedasticity (ARCH) model, the Diffusion Process (DP) model, and the Detrend Fluctuation Analysis (DFA) model of the fractal time series technique. An...
修改评论