Super-resolution Reconstrction Of Image Sequences Based On Interpolation Inverse Projection Iteration

Image has been a main way of expressing and delivering information in our daily life. The image's quality can directly affect the accuracy of delivering informatin, therefore, the imaging equipment of high resolution and high accuracy have emerged in the market. However, the hardware equipment continuous improved, the cost also enhanced continuously.In order to save the cost,meanwhile the imaging effect can be well inproved,this dissertation uses a method of software called super-resolution technology to improve the image's resolution, which will not change the performance of original imaging system.This dissertation's main object is many low-resolution's images, which are in the continuous time for the same object by camera, i.e. image sequences. Based on them, we can reconstruct an image of high resolution and high quality, which is super-resolution reconstructing for image sequences.In order to enhance the resolution of image. Firstly, this thesis uses a method based on the taylor series expansion which proposed by Keren to estimate motion of image registration for image sequences, and use the experiments to prove the algorithm is very effective. Secondly, this dissertation combines the technology of image interpolation into the iterative process of inverse projection, i.e. the method based on the interpolation inverse projection's iteration, which can complete the job of super-resolution reconstruction, and the experiments prove that this algorithm is feasible. Based on this algorithm, In order to reduce the algorithm's iterative times and increase the algorithm's efficiency, this dissertation proposed a fast iteration method which is based on the above algorithm and uses wavelet change to renew the initial value of algorithm, and the experiment proves that, the improved method is correct. Lastly, this dissertation makes a quality evaluation on subject and object for the image which is from the experiments.