【摘要】 把正则图象的误差分成两部分 :一部分为逼近误差 ,它是和原图象相关的确定型误差 ,另一部分为噪声传递误差 ,它是和噪声相关的随机型误差 .通过正则图象代替原图象 ,求得逼近误差的近似值 ,然后将逼近误差近似值补给正则解 ,得到改进型正则解 .对正则解以及逼近误差的计算 ,提出了快速算法 ,并有严密的理论推导 .实验结果表明改进型正则图象的恢复效果优于传统正则解 .在一般的正则化方法中 ,可以借鉴本文提到的方法改进正则解更多还原

【Abstract】 This paper proposes a new technique to improve regularizatization image,It divides, the error of regularized image into two parts: one is approximation error which is definite error related to original image the other is noise propagation error which is random error related to noise.After regularized image takes the place of original image in the express of approximation error,we obtain the approximate value of approximation error which is then compensated to regularized solution.In the case of the computation of regularized solution and approximation errora, fast algorithm is provided with theoretical deducing.Experimental results show that the restoration effect of improved regularized solution is superior to that of general regularized solution.The method in this paper may be used for reference to improve regularized solution in general regularization theory.更多还原