本篇論文是針對加權式最小平方法在記憶體空間和計算時間上提出改善,
並討論有關權值的選取及處理視窗的大小所造成的影響.除此之外,本篇論
文還提出另一種用遺傳基因法則的方法,來修正影像上的節點,使得影像的
相位值連續,並保有原有地形的資訊;相位值連續,則相位解套就可以用路
徑積分的方式輕鬆地解決.在末節還會針對以上的加權式最小平方法,遺傳
基因法,及傳統上的枝節法做模擬比較.並綜合以上的優劣結果,將其應用
在實際上的地形資料上,來求得實際地形的全相位圖.

This thesis proposes some improvement in unwrapping method of
weighted least -square(WLS) , which need huge memory size and
calculating time . Discussion about the weightings and minimal
processing window size will be hold. Besides genetic algorithm
(GA) was proposed to fix the inconsistencies of image that was
called "residues" in branch-cuts method,so that make whole
terrain continuous, and phase unwrapping will easy to go . Three
methods:WLS, GA and branch-cuts method using average algorithm
to processing bad bits will be examined to make a comparison on
simulated terrain . Finally , combine the good of each method
and take it to unwrap the whole real terain .