基于图像内容的多目标量子图像提取与识别关键技术研究

Face to the difficulties of multi-dimensional multi-target color image in the aspect of extraction and recognition based on image content, the project aims to improve the operation speed of image processing by using the parallelism of quantum computing.Firstly,based on the one to one corresponding relationship between pixels and angle of image's space point, employing quantum limit storage expression of multi-dimensional multi-target color image,extracting the image by using repeatedly measurement,the measurement result is converted to Bernouli random variables. According to the Law of large Numbers and Central Limit Theorem, the maximum number of measurements that the single quantum image to be extracted needs can be calculated. Secondly,establishing the connected state of two quantum pure states of image and executing the corresponding quantum logic operation on connecting state, the pixel difference of two images with the same size in corresponding position can be obtained by using repeatedly measurement. Calculate the similarity of the two images so that the two images can be identified. Finally,the extraction and recognition of content-based quantum image can be verified by using Matlab. This project hope to provide a novel idea for traditional digital image processing and pattern recognition and a new way to solve the existing problems.It also has potential practical value and significance in the related application fields of multi-dimensional color image.

针对多维多目标彩色图像在基于图像内容的提取与识别方面的困难,本项目拟利用量子计算的并行特性提高图像处理的运算速度，在图像空间点的像素与角度一一对应关系的基础上,首先采用多维多目标彩色图像的量子极限存储表达式，对其进行多次测量提取出存储的图像，并把测量结果转化为贝努利随机变量，利用大数定理和中心极限定理，推算出提取单幅量子图像需要的最大测量次数；然后建立两个量子图像纯态新的连接态，对连接态进行相应的量子逻辑操作，经过多次测量可以得出两幅大小相同的图像对应位置的像素差，计算出两幅图像相似度的识别函数，从而识别出两幅图像；最后用Matlab对基于图像内容的量子图像提取与识别过程进行仿真验证。希望对传统数字图像处理和模式识别等领域的研究提供一种新的思路和解决现存困难的一种新办法，对多维多目标彩色图像的相关应用领域具有一定的实用价值和指导意义。

针对多维多目标彩色图像在基于图像内容的提取与识别方面的困难,本项目拟利用量子计算的并行特性提高图像处理的运算速度，首先设计了相关相位的规范任意叠加态(Normal Arbitrary Superposition State with Relative Phases， NASSRP)表示多维彩色图像和图像的附加信息，并介绍了基于NASSRP的多维图像存储与检索技术；其次，成功提出和改善量子图像匹配的量子线路设计，量子图像匹配是将量子计算与经典计算机图像匹配技术相结合，利用基本的量子比特门，以及模块化的量子线路来实现量子图像匹配，匹配效率高，适用范围广；最后把量子图像测量问题转化为一个贝努利随机变量测量的正太分布问题，成功实现了量子图像的提取所需的最大次数的理论推导，解决了量子测量盲目性问题，为量子态准备和复杂性估计提供了直接依据。解决了量子图像提取最大测量次数这个Open问题。用Matlab对基于图像内容的量子图像提取与识别过程进行仿真验证。希望对传统数字图像处理和模式识别等领域的研究提供一种新的思路和解决现存困难的一种新办法，对多维多目标彩色图像的相关应用领域具有一定的实用价值和指导意义。