Super-resolution Reconstruction Technology Research Based On Details

Image super-resolution reconstruction is an important aspect of the digital imageprocess technique.It refers to the image processing process that the high-quality andhigh-resolution image could be obtained from single or several low-quality andlow-resolution image by using of the digital signal processing methods. It has low costand, most importantly, the huge amount of the obtained information of the images couldbe made good use of. Also, because of the loosely required property of the images thatthe technique processes, there is no more requirements of the image acquisition system.Given these distinctive advantages, more and more attentions are now being poured intothis field, and it has been investigated extensively.Usually, there exists a wealth of information in the objective world scene. It isdifficult to obtain all the information of the original scene, limited by the imageacquisition system for digital image by the imaging conditions and limitations ofimaging modalities. Especially, the resolution limitation of the acquisition device willresult in the loss of high-frequency image details. The Super-resolution reconstructiontechnique is to try to enhance the image quality of restoration by restoring the “lost”high-frequency information of the image details which is beyond the cut-off frequency.Based on the traditional super-resolution image reconstruction algorithm, a newframe is proposed which concerns several aspects information, the image interpolationreconstruction of single frame to a single frame, the super-resolution imagereconstruction of multi-frame image to single frame, the key techniques ofsuper-resolution image reconstruction and image fusion. The main aspects of thecontribution of this thesis are as follows:1. The theoretical basic and several kinds of model of the super-resolutionreconstruction algorithm is analyzed in great detail, mathematically. A clear descriptionof the super-resolution algorithm principles and calculation methods of single imageinterpolation algorithm and multi-frame image sequence is also given.2. This algorithm is based on the information awareness of the difference betweeneach frame of the image sequence. Based on the non-redundant information isolated from the frame sequence, the redundant high-resolution images and non-redundanthigh-resolution images is obtained respectively by using the traditional super-resolutionreconstruction algorithm. An image with higher resolution than the imaging systems isobtained after the image fusion. Simulation results show that the result obtained by thismethod has a lower root mean square error, and could well maintain the high-frequencydetails of the space.