def applyPredThresh(pixels):
# zero removal for quantile computation
nonzp = pixels[np.any(pixels, 1)]
counts = {c:np.bincount(nonzp[:,c], minlength=256) for c in colors}
qs = {c :
countQuantiles(counts[c], iles) for c in colors}
# clunky:
qDict = {"R8D": qs[R][0], "R9D": qs[R][1],
"G8D": qs[G][0], "G9D": qs[G][1],
"B8D": qs[B][0], "B9D": qs[B][1]}
predictedThreshes = \
dict((c, applyCurrentLM(qDict, c)) for c in colorNames)
# print ",".join("%5.4f" % t for t in predictedThresholds.values())
thresholdNDArray(pixels, predictedThreshes, dropSaturated=True)
def applyPredThresh(pixels):
# zero removal for quantile computation
nonzp = pixels[np.any(pixels, 1)]
counts = {c: np.bincount(nonzp[:, c], minlength=256) for c in colors}
qs = {c: countQuantiles(counts[c], iles) for c in colors}
# clunky:
qDict = {
"R8D": qs[R][0],
"R9D": qs[R][1],
"G8D": qs[G][0],
"G9D": qs[G][1],
"B8D": qs[B][0],
"B9D": qs[B][1]
}
predictedThreshes = \
dict((c, applyCurrentLM(qDict, c)) for c in colorNames)
# print ",".join("%5.4f" % t for t in predictedThresholds.values())
thresholdNDArray(pixels, predictedThreshes, dropSaturated=True)
predictedThreshes = \
dict((c, applyCurrentLM(qDict, c)) for c in colorNames)
# print ",".join("%5.4f" % t for t in predictedThresholds.values())
thresholdNDArray(pixels, predictedThreshes, dropSaturated=True)
origFilename = osp.join(imageDataPath,
"Endosomes/10minend/ser3/10m60xendser31.TIF")
expThreshes = {"R": 17, "G": 41, "B": 34}
currentImage = Image.open(origFilename)
# get base arrays
asArray = {}
asArray['exp'] = fromimage(currentImage).reshape((numImagePoints, 3))
asArray['pred'] = asArray['exp'].copy()
# apply thresholds
thresholdNDArray(asArray['exp'], expThreshes, dropSaturated=True)
applyPredThresh(asArray['pred'])
# reconstruct images and write out
outputPath = "/home/mfenner/scipy_prep/final-images"
for name, arr in asArray.items():
# outImage = Image.merge(currentImage.mode, (arr[R], arr[G], arr[B]))
outImage = Image.fromarray(arr.astype(np.uint8).reshape((1024, 1024, 3)))
outImage.save(
osp.join(outputPath, "10minEndosomeSeries3Slice1-" + name + ".tif"))
print cnd
stackArray = np.empty((0,3), np.uint8)
for threshes in thisCndSet:
print threshes["Slice"],
expertThresholds = dict(((c, threshes[c]) for c in colorNames))
#
# get file->image and thresholds together
# apply threshold
#
exampleFilename = fileNameFormat[threshes["Organelle"]] % threshes
currentImage = Image.open(imageDataPath+exampleFilename)
asArray2 = fromimage(currentImage).reshape((numImagePoints,3))
thresholdNDArray(asArray2, expertThresholds)
stackArray = np.concatenate((stackArray, asArray2))
# zero removal
currAllPixelCt = stackArray.shape[0]
stackArray = stackArray[np.any(stackArray, 1)]
currNonZeroPixelCt = stackArray.shape[0]
print "(%d --> %d)" % (currAllPixelCt, currNonZeroPixelCt)
allPixelCt += currAllPixelCt
nonZeroPixelCt += currNonZeroPixelCt
print "%15s: %12s" % ("with [0,0,0]", allPixelCt)
{"15m60xendser301.TIF",
"15m60xendser301.TIF",
"15m60xendser301.TIF",
"120m60xac17ser24.TIF",
"120m60xac17ser27.TIF"}:
print ".....skipping:"
continue
currentImage = Image.open(imageDataPath+exampleFilename)
pixels = fromimage(currentImage).reshape((numImagePoints,3))
# zero removal
if REMOVE_ZEROS:
pixels = pixels[np.any(pixels, 1)]
counts = {c:np.bincount(pixels[:,c], minlength=256) for c in colors}
qs = {c :
countQuantiles(counts[c], iles) for c in colors}
# import pdb; pdb.set_trace()
# clunky:
qDict = {"R8D": qs[R][0], "R9D": qs[R][1],
"G8D": qs[G][0], "G9D": qs[G][1],
"B8D": qs[B][0], "B9D": qs[B][1]}
predictedThresholds = \
dict((c, applyCurrentLM(qDict, c)) for c in colorNames)
print ",".join("%5.4f" % t for t in predictedThresholds.values())
thresholdNDArray(pixels, predictedThresholds)
stackArray = np.concatenate((stackArray, pixels))
predictedThreshes = \
dict((c, applyCurrentLM(qDict, c)) for c in colorNames)
# print ",".join("%5.4f" % t for t in predictedThresholds.values())
thresholdNDArray(pixels, predictedThreshes, dropSaturated=True)
origFilename = osp.join(imageDataPath,
"Endosomes/10minend/ser3/10m60xendser31.TIF")
expThreshes = {"R":17, "G":41, "B":34}
currentImage = Image.open(origFilename)
# get base arrays
asArray = {}
asArray['exp'] = fromimage(currentImage).reshape((numImagePoints,3))
asArray['pred'] = asArray['exp'].copy()
# apply thresholds
thresholdNDArray(asArray['exp'], expThreshes, dropSaturated=True)
applyPredThresh(asArray['pred'])
# reconstruct images and write out
outputPath = "/home/mfenner/scipy_prep/final-images"
for name, arr in asArray.items():
# outImage = Image.merge(currentImage.mode, (arr[R], arr[G], arr[B]))
outImage = Image.fromarray(arr.astype(np.uint8).reshape((1024,1024,3)))
outImage.save(osp.join(outputPath,
"10minEndosomeSeries3Slice1-"+name+".tif"))
print cnd
stackArray = np.empty((0, 3), np.uint8)
for threshes in thisCndSet:
print threshes["Slice"],
expertThresholds = dict(((c, threshes[c]) for c in colorNames))
#
# get file->image and thresholds together
# apply threshold
#
exampleFilename = fileNameFormat[threshes["Organelle"]] % threshes
currentImage = Image.open(imageDataPath + exampleFilename)
asArray2 = fromimage(currentImage).reshape((numImagePoints, 3))
thresholdNDArray(asArray2, expertThresholds)
stackArray = np.concatenate((stackArray, asArray2))
# zero removal
currAllPixelCt = stackArray.shape[0]
stackArray = stackArray[np.any(stackArray, 1)]
currNonZeroPixelCt = stackArray.shape[0]
print "(%d --> %d)" % (currAllPixelCt, currNonZeroPixelCt)
allPixelCt += currAllPixelCt
nonZeroPixelCt += currNonZeroPixelCt
print "%15s: %12s" % ("with [0,0,0]", allPixelCt)
print "%15s: %12s" % ("removed [0,0,0]", nonZeroPixelCt)
"120m60xac17ser24.TIF",
"120m60xac17ser27.TIF"}:
print ".....skipping:"
continue
currentImage = Image.open(imageDataPath + exampleFilename)
pixels = fromimage(currentImage).reshape((numImagePoints, 3))
# zero removal
if REMOVE_ZEROS:
pixels = pixels[np.any(pixels, 1)]
counts = {c: np.bincount(pixels[:, c], minlength=256) for c in colors}
qs = {c: countQuantiles(counts[c], iles) for c in colors}
# import pdb; pdb.set_trace()
# clunky:
qDict = {
"R8D": qs[R][0],
"R9D": qs[R][1],
"G8D": qs[G][0],
"G9D": qs[G][1],
"B8D": qs[B][0],
"B9D": qs[B][1]
}
predictedThresholds = \
dict((c, applyCurrentLM(qDict, c)) for c in colorNames)
print ",".join("%5.4f" % t for t in predictedThresholds.values())
thresholdNDArray(pixels, predictedThresholds)
stackArray = np.concatenate((stackArray, pixels))
if applyPredThresholds:
# zero removal for quantile computation
pixels = pixels[np.any(pixels, 1)]
counts = {c:np.bincount(pixels[:,c],
minlength=256) for c in colors}
qs = {c:countQuantiles(counts[c], iles) for c in colors}
# clunky:
qDict = {"R8D": qs[R][0], "R9D": qs[R][1],
"G8D": qs[G][0], "G9D": qs[G][1],
"B8D": qs[B][0], "B9D": qs[B][1]}
predThreshes = dict((c, applyCurrentLM(qDict,
c)) for c in colorNames)
# print ",".join("%5.4f" % t for t in predThresholds.values())
thresholdNDArray(pixels, predThreshes, dropSaturated=True)
expArray = np.concatenate((expArray, pixels))
#########################################
# all to here is necessary to get joined series
#########################################
org = simplifyOrgStain(cnd["Organelle"], cnd["Stain"])
t = timeToIdx(cnd["Time"])
print org, t
# convert image stack to counts and add to histograms
for c1, c2 in colorPairs:
probs = toProbs(expArray[:,c1], expArray[:,c2],
removeNonresponders = removeNonresponders,
