Refs #114. autocrop: discard influence of the borders.

This problem arises when a particular scan of CT has some stripes at the
bottom left of the images. Hassina said it is the table.
That stripes throw the autocrop off.
This hack just remove the borders from the consideration of autocropping.

python/imars3d/autocrop.py

@@ -1,6 +1,6 @@
 import numpy as np
 
-def calculateCropWindow(series):
+def calculateCropWindow(series, normalize=False):
     # estimate average
     ave = estimateAverage(series)
     from . import io
@@ -12,7 +12,10 @@ def save(d, p):
     sm = ndimage.median_filter(ave, 9) # 21)
     save(sm, "sm-estimate-ave.tiff")
     # make sure all numbers are smaller than 1.
-    max = np.max(sm);  sm/=max
+    if normalize:
+        max = np.max(sm);  sm/=max
+    else:
+        sm[sm>1] = 1
     # get background intensities
     left = sm[:, :10];    left_median = np.median(left)
     top = sm[:10, :];     top_medain = np.median(top)
@@ -28,6 +31,9 @@ def save(d, p):
         xmin = np.min(X1); xmax = np.max(X1)
         ymin = np.min(Y1); ymax = np.max(Y1)
         return xmin, xmax, ymin, ymax
+    # there might be outliers in borders, just clear the borders out
+    sm[:, :25] = background_int;   sm[:25, :] = background_int
+    sm[:, -25:] = background_int;   sm[-25:, :] = background_int
     # find rectangle for real data
     Y, X = np.where(sm < background_int*0.95)
     ymax = Y.max(); ymin = Y.min()