A METHOD, AN APPARATUS AND A COMPUTER-READABLE MEDIUM FOR PROCESSING A NIGHT VISION IMAGE DATASET

摘要

A kind of method of quality for strengthens view data collection is disclosed, for example, reducing the noise that noisy image data is concentrated and increasing the contrast of image data concentration. This method can be used for handling the sequence of image data set, such as. Night vision image data set, wherein the sequence includes at least two image data sets, and each image data set has at least two pixels, and wherein each pixel has intensity value. This method includes the structure tensor for calculating each pixel that the image data for including is concentrated in image data set sequence; Structure tensor based on each pixel in described image data set calculates the value in summation core; It calculates in the first image data set, the value in the summation core is used as weight; Using the weighed intensities value of each pixel in described image data set as treated output image data concentrate each respective pixel processing intensity value; The rotation local coordinate system that wherein summation core is DESCR; The result shows that the reference axis of local coordinate system is consistent with direction coordinate. About the feature vector of the structure tensor, wherein described eigenvector describes in the global coordinate system of image data set, and scales the reference axis for the local coordinate system that wherein summation core is described by amount relevant to the characteristic value of structure. Pass through Width Function W(α i) the CcIn tensors of=Sigi i, wherein the characteristic value depends on the Strength Changes amount on their corresponding feature vector directions, and the Width Function is to reduce function, so that w (0)=max and lima ∞ w=min.