The Study Of The Acquirement And Processing Technology Of High Quality Images

Human obtains information, more than 80% of which is accepted by vision. As the main medium of visual information, images are widely applied in many fields such as spaceflight, medical treatment, security and so on. As technology developing, these industry has stricter demands for image quality. So, it is still a popular hotspot that how to improve the quality of images in image processing. According to the flow of the obtaining of high quality images, this paper is established in image focusing and image denoising, to study how to get a clearer image while collecting and processing.The goal of focus is to obtain the clear images, which is base of high quality image. In this filed, this paper studies existing focusing algorithms and to find that Depth for Focus is more effective than other means. Depth for Focus is a method which collects images with lens, and estimates them with a definition estimation algorithm. The result of the estimation will be used for controlling the rolling of lens to get a clearer image. In this method, the performance of the definition estimation algorithm is the key to the obtaining of high quality images. While, the traditional algorithms are usually limited by their poor precision and adaptability. So, this paper proposes a definition evaluation algorithm for out-of-focus images based on quadtree decomposition, which combines quadtree decomposition and traditional EOG definition evaluation algorithm. It estimates the whole definition by quadtree decomposition, which is also used for resist the noise. At the same time, it estimates the definition in each block of the result of quadtree decomposition to optimize the result. Tested by the contrasting experiment, the enhanced algorithm has a stronger sensitivity and capability of resisting noise.In practice, there will still be some noise in the image got by focusing, which is caused by the interference in and out of the system and is able to cause a pool quality. To improve the quality of image, denoising must be study. After compared with a lot of existing denoising means, NLM is proved to have great performance. So this paper make a study on it but find that it can’t avoid the disadvantage of denoising method, which is the loss of detail. What’s more, it is a limitation that the parameters of NLM should be chosen by human. With the consequence of analyzing its theory, development, advantages and disadvantages, this paper proposes a enhanced Nonlocal Means which can protect the edges and texture well. This algorithm introduces the weight of texture and the edge detection method, adding the weight grads to the pixels on edge. The texture factor make sure that the denoising intensity changes by image, as the edge factor optimize the method which decides the choice of reference point in the edge area. Tested by experiments, the enhanced algorithm has a better objective evaluation and preserves more edges and texture. At the same time, this paper also proposes another enhanced Nonlocal Means to self the problem of self-adaption, in which the search window and parameter can be chosen adaptively. This algorithm achieves the adaptive produce of the search window by noise detection and snake window, which can reduce the bad influence brought by noise pixel. It also determines the filtering parameter based on the standard deviation of noise and the distance of pixels in space. It is proved that this enhanced algorithm improves the denoising capability of traditional Nonlocal Means while achieving self-adaption, and it can processing various noise.