| It can break through the bottleneck of limit of hardware in improving spatial resolution of remote sensing images through transforming the modes of sampling and enhancing the resolution to some extent.In those modes,tilted sampling avoids the problem of displacement accuracy of sub-pixel and just need no more than one linear array,unlike super mode or hyper mode.(1)The formation of aliasing of tilted sampling was gained through analysis of sampling grids and Shannon's sampling theorem,then reciprocal cell was introduced which satisfies Shannon's sampling theorem to compensate frequency spectrum.It proved condition of application of the Reciprocal Cell and an effective approach of antialiasing.(2)The relationship among angles with the push-broom direction,aliasing and effective resolution were lucubrated.On basis of sampling theorem and formulation of aliasing,the measurement of aliasing degree was designed(Aliasing Index),moreover the AI of regular sampling and titlting sampling(from 1° to 89°)were calculated.Thus the angle range of less aliasing was deduced.When the sampling distance was the detector size,the less aliasing angle range was 57° to 89°,when the sampling distance was set less than detector size,the range was 1° to 58° and 79° to 89°,when the sampling distance was set more than detector size,there were no angles,at this situation,the effect of aliasing must be took account in case of reducing the image quality,otherwise titling sampling won't work.As well as Effective Eesolution(ER)was brough in to analyze the resolution,the rule was got through calculating the ER of 1° to 89°(interval is 1°).When the sampling distance was the detector size,the bigger ER of the angle range was 57° to 89°,when the sampling distance was less than the detector size,the bigger ER of the angle ranges were 1° to 58° and 79° to 89°,when the sampling distance was set more than detector size,there were no angle ranges whose ER were much bigger.Based on the above research,the relationship between aliasing and resolution was gained:when the sampling distance was the detector size,the relationship between aliasing and resolution was y =0.0749x40 593(x is independent variable,and y is dependent variable),when the sampling distance was less than the detector size,their relationship was y = 0.0098x2.4473,when the sampling distance was set more than detector size,their relationship was y = 0.3652x2-0.6083x+0.2594.Though the expression was not the same,they all demonstrated the effect of alasing on resolution was more and more huge,when the aliasing was big in a certain extent.Then came to the conclusion of angles of less aliasing than normal sampling,and the angles of the least aliasing;angles of higher resolution than normal sampling and angles of the highest resolution;the 'quantity' relationship between aliasing and resolution.(3)In terms of the ambiguity of the titling angles of oblique mode and brought sawtooth and aliasing once more by correct process,optimal angle(27°)was found that the sampling interval in horizonal and vertical direction are the same.With this angle,image can keep geometrical features of scenery distortionless,moreover the grid or geometrical correction is unnecessary.So the experiments of images collected with the angle of 27° were designed,and the images were restored using Reciprocal Cell.The evaluation of resolution according to MTF showed that the resolution of restored images using Reciprocal Cell were two times higher than original images.(4)Blurring model(two-dimension modulation transfer function)was estimated according to the analysis of imaging model.In practice,one-dimension MTF was estimated used knife edge and the target was self-made,furthermore,two-dimension MTF was compounded as bluring model by vertical and horizonal one-dimension MTF.The ill-posedness problem can be solved by the blurring model during restoration process.Combining bluring model and Reciprocal Cell and restored methods of corrected reversal,Weiner and TV,then the advantages of details,definition,amount of information and resolution was Weiner through assessment indexes of definition,amount of information,resolution,signal to noise ratio,total spatial liveness.(5)In terms of transformed noise during restoration and large noise in images,statistic method was used to judge the type of noise according to study the models of noise at first,then with respect to larger Gaussian noise,blend of context model and scale space method in wavelet domain were brought in to denoise.However,the sequence of denosing,deblurring,dealiasing was ordered.Finally,the images sampled in angle of 27° with large Gaussian noise were restored.The evaluation system showed that the method in this paper was good for reducing aliasing,removing mass noise to make restored image and improving spatial resolution. |
PDF Full Text Request