First implementation of Integer Up/Downsampling

skimage/transform/rescale.py

@@ -0,0 +1,51 @@
+# TODO : Doc, Tests, PEP8 check
+
+import numpy as np
+
+def downsample(image, factors, how = 'sum'):
+
+	is0 = image.shape[0]
+	is1 = image.shape[1]
+	f0 = factors[0]
+	f1 = factors[1]
+
+	if (f0 - int(f0) != 0) or (f1 - int(f1) != 0):
+		return "Use resample for non-integer downsampling"
+	cropped = image[: is0 - (is0 % f0), : is1 - (is1 % f1)]
+	out = np.zeros((cropped.shape[0] / f0, cropped.shape[1] / f1))
+
+	if how == 'sum':
+		for i in range(cropped.shape[0]):
+			for j in range(cropped.shape[1]):
+				out[int(i / f0)][int(j / f1)] += cropped[i][j]
+		return out
+
+	if how == 'mean':
+		for i in range(cropped.shape[0]):
+			for j in range(cropped.shape[1]):
+				out[int(i / f0)][int(j / f1)] += cropped[i][j] / float(f0 * f1)
+		return out
+
+
+def upsample(image, factors, how = 'divide'):
+
+	is0 = image.shape[0]
+	is1 = image.shape[1]
+	f0 = factors[0]
+	f1 = factors[1]
+
+	if (f0 - int(f0) != 0) or (f1 - int(f1) != 0):
+		return "Use resample for non-integer upsampling"
+	out = np.zeros((f0 * image.shape[0], f1 * image.shape[1]))
+
+	if how == 'divide':
+		for i in range(out.shape[0]):
+			for j in range(out.shape[1]):
+				out[i][j] = (image[i / f0][j / f1]) / float(f0 * f1)
+		return out
+
+	if how == 'uniform':
+		for i in range(out.shape[0]):
+			for j in range(out.shape[1]):
+				out[i][j] = (image[i / f0][j / f1])
+		return out