應用廣義加權平均集成運算於影像壞點偵測

Abstract

本論文應用廣義加權平均建立的區間值模糊關係進行灰階影像壞點偵測。首先，我們使用兩個加權參數，對整張影像以 視窗內中心像素和其八鄰域像素進行加權平均集成運算。接著，為了避免單一組大的差值掩蓋其它組差值呈現，我們應用飽和門檻值轉換函數(Saturation threshold transfer function)在像素差值上。最後，經由門檻值(Threshold)作用後，可獲得影像壞點較合理的估測。 為了降低壞點誤判的個數，我們藉著離散型梯度演算法的概念進行加權平均參數的學習。另外，為了修正影像更高的訊雜比，相對於好點誤判，我們加重壞點誤判的權重約20-100倍。除此之外，我們更提出兩個不同的訓練階段，以維持影像的銳利度。從四張添加脈衝雜訊的灰階合成影像訓練結果顯示，整合區間值模糊關係與加權平均差值集成演算法，能產生更為強健的壞點偵測。

In this thesis, we apply weighted generalized mean to construct interval-valued fuzzy relations for grayscale image noise detection. First, we employ two weighting parameters and perform the weighted mean aggregation for the central pixel and its eight neighbor pixels in a sliding window across the image. Then, in order to counter the over-weighting of a big difference term, we apply a saturation threshold transfer function to pixel difference values. Finally, the image noise map is obtained through a threshold operation. In order to decrease the noise detection error, weighting parameters of the mean can be learned by the gradient method caste in discrete formulation. Moreover, to get higher PSNR in the corrected image, we have, in the training, put multiple weight ranging from 20 to 100, on erroneous noisy than that on the erroneous non-noise pixel. Besides, we also propose two training stages for the purpose of maintaining image sharpness and correction. By the training results of four grayscale natural images with adding impulse noises, we have shown that the integration of interval-valued fuzzy relations with the weighted mean aggregation algorithm can effectively detect the image noise and do the correction hereafter.