Study On Local Image Reconstructing Algorithm Based On The Influence Function Between Point And Source

A good visual effect has been the main target of present researchers in image reconstruction,and the accurate local areas reconstruction has been neglected. However,with the development ofimage analyzation and recognition, more and more reconstructed images have been used for featureextraction. Therefore, accurate image reconstruction is arousing researchers’ attention.At the sametime, the reconstruction efficiency is also being taken seriously because of the real-time requirementand video processing.Point to the two problems above, stable field model of image local area was proposed. Basedon the fact that it was the stable result of the interaction between surface texture and structure ofobject and light,a stationary image is regarded as a steady energy field in this model. And then,asteady field equation of the image is set up based on electrostatic field model. During thereconstruction process,the lost texture would be acquired by solving the equation according toinitial condition and boundary condition, and the known pixel information is spread into the lostarea iteratively. Experimental results show that this model not just has a high efficiency butachieves preliminary results in accurate reconstruction.Based on the stable model, a new reconstructing model of image local region based on thepoint-source influence function is proposed. Also it describes the local region of an image as asteady field and a equation is set up. In this model the missing pixel is the “point” and the known isthe “source”. The proposed aims at reconstructing each missing pixel accurately by calculating theinfluence of the known pixels around; According to the directional derivative of the stable field, amethod of calculating the influence function is determined by analyzing both the similarities anddifferences between the missing pixel and the known, so that the reconstruction model could beachieved. Experimental results show that the proposed not just can reconstruct the edges and textureof damaged images clearly and have a fairly higher efficient, but also it keep a fairly good visualeffect.In the research of the reconstruction model, the gradient has an important influence in theeffect of the edge reconstruction. After analyzing the calculation method of the gradient,a newcalculation mode is proposed. In this mode, firstly, the unidirectional calculation of gradientcomponent is used to avoid the edge fuzziness and reduce the probability of involving the lostpixels in gradient calculation. Secondly, the “nearly principle” is used to set prioritie for knownsurrounded pixels, that is the known pixel which is more adjacent to the lost one will be chosen firstly, it ensures the effectively reflecting of edges in the gradient. The proposed calculation modelis proved to be effective in improving the reconstructing effect of edges by experimental results.