在最小二乘匹配法中，非常注重影像間輻射分布情形，但是當影像的來源不同時，兩影像間輻射分布情況差異極大，為了在不同來源影像上進行最小二乘匹配，勢必要修正影像窗，使兩者輻射分布類似。 本研究中將目標窗影像分塊，並以分塊後之結果計算個別的輻射參數以修正目標窗，並利用修正後目標窗與搜尋窗，進行最小二乘匹配。結果顯示這個方法是可行的，可使原本最小二乘匹配法無法匹配的點成功匹配，且匹配成功與匹配正確的點數都增加。 研究中以目標窗影像和搜尋窗影像的分塊成果當作方差分量之分組依據，以最佳線性無偏估計式( Blues )求解方差分量，求解權矩陣代入最小二乘匹配法中計算。最後的計算結果顯示，加入方差分量的結果可增加匹配成功及正確的點數，以及微量提升匹配的精度。 In least squares matching (LSM), image radiometric distribution is very important. But the radiometric distributions differ from different source images. In order to match images by LSM from different sources, it must modulate images windows for making similar radiometric distributions. In this study, we segment target windows to get individual radiometric parameters for segmentations, and modulate target window. Using modulated target window to match searching window by LSM is a practicable method from experiment result. The points can’t be matched are matched by this method, and the numbers of matching success rate and matching correct are increasing rate. We use search window and target window segments to decide the class of variance component in the research. We use Best linear unbiased estimators ( Blues ) to calculate variance component, calculate weight matrix for LSM. The results show that after weighting by variance component, the points of matching success and matching correct are increasing and the matching accuracy can increase little.
