def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(img.shape[2]):
outputBytes.append(img[y, x, c])
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
outputFile.write('%d %d %d\n' % (img.shape[0], img.shape[1], img.shape[2]))
outputFile.write(outputBytes)
# Print information about the compression
inSize = img.shape[0] * img.shape[1] * img.shape[2]
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
startTime = time.time()
outputBytes = bytearray()
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(img.shape[2]):
outputBytes.append(img[y, x, c])
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
outputFile.write('%d %d %d\n' % (img.shape[0], img.shape[1], img.shape[2]))
outputFile.write(outputBytes)
# Print information about the compression
inSize = img.shape[0] * img.shape[1] * img.shape[2]
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
# outputBytes = []
outputBytes = bytearray()
diff = []
# it's a single-channel image
if len(img.shape) == 2:
for y in range(img.shape[0]):
for x in range(img.shape[1]):
# Predictive encoding
prediction = int(
img[y - 1,
x]) + (int(img[y, x - 1]) - int(img[y - 1, x - 1])) / 2
diff.append(int(img[y, x]) - int(prediction))
else: # it's a multi-channel image
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(img.shape[2]):
# Predictive encoding
prediction = int(img[y - 1, x, c]) + (
int(img[y, x - 1, c]) - int(img[y - 1, x - 1, c])) / 2
diff.append(int(img[y, x, c]) - int(prediction))
# print(diff)
# construct initial dictionary
dict_size = 256
dictionary = {
struct.pack("h", i): i + 256
for i in xrange(-dict_size, dict_size)
}
# LZW encode the diff array
s = ''
for i in range(len(diff)):
x = struct.pack('h', diff[i])
temp = s + x
if temp in dictionary:
s = temp
else:
s_in = dictionary[s]
s1 = s_in >> 8
s2 = s_in & 255
outputBytes.append(s1)
outputBytes.append(s2)
if len(dictionary) < 65536:
dictionary[temp] = dict_size
dict_size += 1
s = x
# encode the last s
s_in = dictionary[s]
s1 = s_in >> 8
s2 = s_in & 255
outputBytes.append(s1)
outputBytes.append(s2)
# print(dictionary)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
outputFile.write('%d %d %d\n' % (img.shape[0], img.shape[1], img.shape[2]))
outputFile.write(str(outputBytes))
# Print information about the compression
inSize = img.shape[0] * img.shape[1] * img.shape[2]
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of components in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
# initialize dictionary
d = {}
counter = 256
for i in range(-counter, counter):
d[str(i)] = i
# Set Dictionary limit
# Make a list to hold bytes
tempBytes = []
# A counter for the number of bytes
numBytes = 0
multichannel = True
# for a single channel image
if (len(img.shape) == 2):
multichannel = False
# Go through whole image
for y in range(img.shape[0]):
for x in range(img.shape[1]):
# Initialize prediction to image value
prediction = img[y][x]
#"""
# Modify prediction to show the difference between prior pixels and current pixel
if (x != 0):
prediction = prediction - img[y][x - 1]
elif (y != 0):
prediction = prediction - img[y - 1][x]
else:
prediction = prediction - (img[y][x - 1] / 3 +
img[y - 1][x] / 3 +
img[y - 1][x - 1] / 3)
#"""
# Add the predicted value to the bytestream
tempBytes.append(prediction)
numBytes += 1
# Multi-Channel
else:
# Go through whole image and channels
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(img.shape[2]):
# Do predictive encoding
# Initialize prediction to image value
prediction = img[y][x][c]
# Modify prediction to show the difference between prior pixels and current pixel
#""""
if (x != 0):
prediction = prediction - img[y][x - 1][c]
elif (y != 0):
prediction = prediction - img[y - 1][x][c]
else:
prediction = prediction - (img[y][x - 1][c] / 3 +
img[y - 1][x][c] / 3 +
img[y - 1][x - 1][c] / 3)
#""""
# Add the predicted value to the bytestream
tempBytes.append(prediction)
numBytes += 1
# Using a string variable as it allows for concatenation
s = ""
# Set s to the first value of the bytestream
s = str(tempBytes[0])
# Go through all bytes
for i in range(1, numBytes):
# Do LZW encoding
# If trying to add entry larger than max size of the dictionary reinitialize the dictionary
if (counter >= twoBytes):
counter = 256
d = {}
for i in range(-counter, counter):
d[str(i)] = i
# Add the next byte to the current string. Uses a delimeter to distinguish numbers
w = s + "|" + str(tempBytes[i])
# Checking if it has been seen before
if w in d:
s = w
else:
# Output bytes by splitting integer into two bytes, this allows for a larger dictionary
outputBytes.append((int(d[s]) >> 8) & 0xFF)
outputBytes.append(int(d[s]) & 0xFF)
# Add to dictionarry
d[w] = counter
counter += 1
s = str(int(tempBytes[i]))
# Check if the last byte was added or not
if s in d:
outputBytes.append((int(d[s]) >> 8) & 0xFF)
outputBytes.append(int(d[s]) & 0xFF)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
if (multichannel):
outputFile.write('%d %d %d\n' %
(img.shape[0], img.shape[1], img.shape[2]))
else:
one = 1
outputFile.write('%d %d %d\n' % (img.shape[0], img.shape[1]), one)
outputFile.write(outputBytes)
# Print information about the compression
if (multichannel):
inSize = img.shape[0] * img.shape[1] * img.shape[2]
else:
inSize = img.shape[0] * img.shape[1]
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
# outputBytes = np.array([],dtype = np.uint16)
tempBytes = []
channels = 0
# -256 - 0 maps to 0 - 256, 1-255 mapps to 257 - 511
baseDict = genDict()
if len(img.shape) == 2:
channels = 1
elif len(img.shape) == 3:
channels = 3
else:
print("unrecognized image format")
return
pArray = []
# if 1 channel, loop through every pixel
if (channels == 1):
for x in range(img.shape[0]):
for y in range(img.shape[1]):
# initial pixel, don't make prediction
if (x == 0 and y == 0):
pArray.append(int(img[0, 0]))
else:
# make prediction using previous pixel
if x == 0:
pArray.append(int(img[x, y]) - int(img[x, y - 1]))
else:
pArray.append(int(img[x, y]) - int(img[x - 1, y]))
# Image has 3 channels, loop through all pixels and channels
else:
for x in range(img.shape[0]):
for y in range(img.shape[1]):
for c in range(img.shape[2]):
# Initial pixel, no prediction
if (x == 0 and y == 0):
pArray.append(int(img[0, 0, c]))
else:
if x == 0:
pArray.append(
int(img[x, y, c]) - int(img[x, y - 1, c]))
else:
pArray.append(
int(img[x, y, c]) - int(img[x - 1, y, c]))
# number of symbols to encode
numSymbols = img.shape[0] * img.shape[1] * channels
# LZW Compression
s = str(pArray[0])
nextDictIndex = len(baseDict)
for i in range(1, numSymbols):
x = str(pArray[i])
if s + "," + x in baseDict:
s = s + "," + x
else:
tempBytes.append(baseDict[s])
baseDict[s + "," + x] = np.uint16(nextDictIndex)
nextDictIndex += 1
s = x
tempBytes.append(baseDict[s])
outputBytes = np.array(tempBytes, dtype=np.uint16)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
stringImgShape = []
for num in img.shape:
stringImgShape.append(str(num))
outputFile.write('%s\n' % headerText)
outputFile.write(' '.join(stringImgShape) + '\n')
outputFile.write(outputBytes)
# Print information about the compression
inSize = numSymbols
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress( inputFile, outputFile ):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread( inputFile ).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
# Initialize dictionary
maxsize = 65536
dict_size = 256
d = initializeDictionary(dict_size)
s = ''
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(img.shape[2]):
fp = 0
if(len(img.shape) > 2):
#fp = getprediction(img,x,y,c,True)
if(x > 0):
fp = img[y,x-1,c]
e = img[y,x,c] - fp
else:
#fp = getprediction(img,x,y,0,False)
if(x > 0):
fp = img[y,x-1]
e = img[y,x] - fp
if(s + chr(e) in d):
s += chr(e)
else:
#sq = convertchar2num(d[s])
sq = d[s]
if(len(sq) < 2):
sq = chr(0) + sq
outputBytes.append(sq[0])
outputBytes.append(sq[1])
d[s + chr(e)] = convertnum2char(dict_size)
dict_size += 1
if(dict_size > maxsize):
dict_size = 256
d = initializeDictionary(dict_size)
s = chr(e)
sq = d[s]
if(len(sq) < 2):
sq = chr(0) + sq
outputBytes.append(sq[0])
outputBytes.append(sq[1])
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write( '%s\n' % headerText )
if len(img.shape) > 2:
outputFile.write( '%d %d %d\n' % (img.shape[0], img.shape[1], img.shape[2]) )
else:
outputFile.write( '%d %d\n' % (img.shape[0], img.shape[1]))
outputFile.write( outputBytes )
# Print information about the compression
if len(img.shape) > 2:
inSize = img.shape[0] * img.shape[1] * img.shape[2]
else:
inSize = img.shape[0]*img.shape[1]
outSize = len(outputBytes)
sys.stderr.write( 'Input size: %d bytes\n' % inSize )
sys.stderr.write( 'Output size: %d bytes\n' % outSize )
sys.stderr.write( 'Compression factor: %.2f\n' % (inSize/float(outSize)) )
sys.stderr.write( 'Compression time: %.2f seconds\n' % (endTime - startTime) )
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
dimensions = len(img.shape)
# if single channel
if dimensions == 2:
# Empty dictionary
d = {}
# Current dictionary index
dict_index = 0
# Previous value
prev = 0
for i in range(-255, 256):
# Stores the binary value
d[(struct.pack('>h', i))] = i + 255
dict_index = i + 255
outputBytes = bytearray()
# p is the byte pattern that will be passed into the dictionary
p = ""
p_c = ""
for y in range(img.shape[0]):
for x in range(img.shape[1]):
# c is the current byte being analyzed
c = int(img[y][x]) - int(prev)
prev = img[y][x]
bc = struct.pack('>h', c)
p_c = p + bc
# if the character c is already in the dictionary add it to the present character stream
if p_c in d:
p = p_c
else:
val = struct.pack('>H', d[p])
b1, b2 = struct.unpack('>BB', val)
outputBytes.append(b1)
outputBytes.append(b2)
if dict_index < 65535:
dict_index += 1
d[p_c] = dict_index
p = bc
outputFile.write('%d %d\n' % (img.shape[0], img.shape[1]))
outputBytes.append(img[y, x])
# if multichannel
elif dimensions > 2:
# Empty dictionary
d = {}
# Current dictionary index
dict_index = 0
# Previous values
prev = []
for i in range(img.shape[2]):
prev.append(0)
for i in range(-255, 256):
# Stores the binary value
d[(struct.pack('>h', i))] = i + 255
dict_index = i + 255
outputBytes = bytearray()
# p is the byte pattern that will be passed into the dictionary
p = ""
p_c = ""
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for channel in range(img.shape[2]):
# c is the current byte being analyzed
c = int(img[y][x][channel]) - int(prev[channel])
prev[channel] = img[y][x][channel]
bc = struct.pack('>h', c)
p_c = p + bc
# if the character c is already in the dictionary add it to the present character stream
if p_c in d:
p = p_c
else:
val = struct.pack('>H', d[p])
b1, b2 = struct.unpack('>BB', val)
outputBytes.append(b1)
outputBytes.append(b2)
if dict_index < 65535:
dict_index += 1
d[p_c] = dict_index
p = bc
outputFile.write('%d %d %d\n' %
(img.shape[0], img.shape[1], img.shape[2]))
outputBytes.append(img[y, x, channel])
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
outputFile.write(outputBytes)
# Print information about the compression
inSize = img.shape[0] * img.shape[1]
outSize = len(outputBytes)
endTime = time.time()
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
diff_image = []
# create diffrence array
if len(img.shape) != 3: # single-channel image
num_of_channels = 1
for y in range(img.shape[0]):
for x in range(img.shape[1]):
if x == 0: # first pixel
diff_image.append(img[y][x])
else:
diff_image.append(img[y][x] - img[y][x - 1])
else: # not single-channel image
num_of_channels = img.shape[2]
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(img.shape[2]): # include colour
if x == 0: # first pixel
diff_image.append(img[y][x][c])
else:
diff_image.append(img[y][x][c] - img[y][x - 1][c])
init_dict = {} # initialize dictionary
dict_size = (255 * 2)
for i in range(dict_size + 1):
init_dict[str(i - 255) +
","] = i # initialize all possible diffrences (no negative)
dict_copy = init_dict.copy()
sub_string = ""
for element in diff_image:
if dict_size >= 65536: # if larger than max allowed size
dict_copy = init_dict.copy()
dict_size = (255 * 2)
sub_string += str(element) + ","
if not sub_string in dict_copy:
dict_size += 1
dict_copy[sub_string] = dict_size
append_key = ",".join(sub_string.split(",")[:-2]) + ","
byte_one = dict_copy[append_key] // (
2**8) # get first byte value (from int)
byte_two = dict_copy[
append_key] & 0b11111111 # get second byte value (from int)
outputBytes.append(byte_one)
outputBytes.append(byte_two)
sub_string = str(element) + ","
byte_one = dict_copy[sub_string] // (2**8)
byte_two = dict_copy[sub_string] & 0b11111111
outputBytes.append(byte_one)
outputBytes.append(byte_two)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
print num_of_channels
outputFile.write('%s\n' % headerText)
outputFile.write('%d %d %d\n' %
(img.shape[0], img.shape[1], num_of_channels))
outputFile.write(outputBytes)
# Print information about the compression
if len(img.shape) == 3:
inSize = img.shape[0] * img.shape[1] * img.shape[2]
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' %
(inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
else:
inSize = img.shape[0] * img.shape[1] * num_of_channels
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' %
(inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
index = 256
# create initial dict for LZW compression containing all possible differences
dictionary = dict((chr(i), i) for i in xrange(index))
difs = []
# loop through each channel of each pixel and compute the predictive difference
for y in range(img.shape[0]):
for x in range(1, img.shape[1]):
for c in range(img.shape[2]):
dif = 0
if (img.shape[2] == 1): # single channel
dif = abs(int(img[y, x]) - int(img[y, x - 1]))
else: # more channels
dif = abs(int(img[y, x, c]) - int(img[y, x - 1, c]))
difs.append(dif)
# for testing to check the similarity between uncompressed differences after decoding
unencoded = open('unencoded.txt', 'w')
for d in difs:
unencoded.write(str(d) + '\n')
unencoded.close()
s = ''
# loop through the differences and perform LZW compression
for x in difs:
sx = s + chr(x)
if sx in dictionary:
s = sx
else:
# turn index into 2 bytes
b = struct.pack('H', dictionary[s])
for byte in b:
outputBytes.append(byte)
# keep dictionary size below 65535 to enforce 2 byte indices
if len(dictionary) < 65535:
dictionary[sx] = index + 1
index += 1
s = chr(x)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
outputFile.write('%d %d %d\n' % (img.shape[0], img.shape[1], img.shape[2]))
outputFile.write(outputBytes)
# Print information about the compression
inSize = img.shape[0] * img.shape[1] * img.shape[2]
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
baseDict = {}
for i in range(512):
baseDict[str(i)] = i
pixelValues = []
if (len(img.shape) == 2):
channels = 1
for x in range(img.shape[0]):
for y in range(img.shape[1]):
pixelValues.append(img[x, y])
elif (len(img.shape) == 3):
channels = 3
for x in range(img.shape[0]):
for y in range(img.shape[1]):
for c in range(img.shape[2]):
pixelValues.append(img[x, y, c])
else:
print("invalid image format, exiting")
quit()
predictions = []
predictions.append(pixelValues.pop(0) + 255)
for i in pixelValues:
predictions.append(int(pixelValues[i]) - int(pixelValues[i - 1]) + 255)
numSymbols = img.shape[0] * img.shape[1] * channels
nextIndex = len(baseDict)
s = ""
for i in range(numSymbols):
x = str(predictions[i])
if s + x in baseDict:
s = s + x
else:
outputBytes.append((int(baseDict[s]) >> 8) & 0xFF)
outputBytes.append(int(baseDict[s]) & 0xFF)
baseDict[s + x] = nextIndex
nextIndex += 1
s = x
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
if (channels == 1):
outputFile.write('%d %d\n' % (img.shape[0], img.shape[1]))
else:
outputFile.write('%d %d %d\n' %
(img.shape[0], img.shape[1], img.shape[2]))
outputFile.write(outputBytes)
# Print information about the compression
inSize = img.shape[0] * img.shape[1] * channels
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress(inputFile, outputFile):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread(inputFile).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
#
# Note that single-channel images will have a 'shape' with only two
# components: the y dimensions and the x dimension. So you will
# have to detect this and set the number of channels accordingly.
# Furthermore, single-channel images must be indexed as img[y,x]
# instead of img[y,x,1]. You'll need two pieces of similar code:
# one piece for the single-channel case and one piece for the
# multi-channel case.
startTime = time.time()
outputBytes = bytearray()
# detect shape and set channels accordingly
if len(img.shape) == 2:
channels = [img.flatten()]
elif len(img.shape) == 3:
channels = [img[:, :, i].flatten() for i in range(img.shape[2])]
# loop through each channel to encode each individually
for chnl in channels:
lzwD, i = initLZWD()
#initialize counters to keep track of previous value and previous sequence for LZW
prevP = 0
prevS = ()
for p in chnl:
pDiff = p - prevP
currS = prevS + (pDiff, )
# if sequence not in dict, write to stream
if currS not in lzwD:
# stop at max 16 bit size -1 = 65535
if i < 65535:
lzwD[currS] = i
i += 1
# get value that was prevS
val = lzwD[prevS]
# encode into bytes
smallB = val % 256
bigB = val / 256
# append bytes to output
outputBytes.append(bigB)
outputBytes.append(smallB)
# reset prevS to new previous sequence
prevS = (pDiff, )
# sequence in dict, don't write new sequence, add to existing sequence
else:
# onto next sequence
prevS = currS
prevP = p
# finish LZW by encoding end data
if len(prevS) > 1:
# same as 'if sequence not in dict'
val = lzwD[prevS]
smallB = val % 256
bigB = val / 256
outputBytes.append(bigB)
outputBytes.append(smallB)
# end flags
outputBytes.append(255)
outputBytes.append(255)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write('%s\n' % headerText)
outputFile.write('%d %d %d\n' %
(img.shape[0], img.shape[1], len(channels)))
outputFile.write(outputBytes)
# Print information about the compression
inSize = img.shape[0] * img.shape[1] * len(channels)
outSize = len(outputBytes)
sys.stderr.write('Input size: %d bytes\n' % inSize)
sys.stderr.write('Output size: %d bytes\n' % outSize)
sys.stderr.write('Compression factor: %.2f\n' % (inSize / float(outSize)))
sys.stderr.write('Compression time: %.2f seconds\n' %
(endTime - startTime))
def compress( inputFile, outputFile ):
# Read the input file into a numpy array of 8-bit values
#
# The img.shape is a 3-type with rows,columns,channels, where
# channels is the number of component in each pixel. The img.dtype
# is 'uint8', meaning that each component is an 8-bit unsigned
# integer.
img = netpbm.imread( inputFile ).astype('uint8')
# Compress the image
#
# REPLACE THIS WITH YOUR OWN CODE TO FILL THE 'outputBytes' ARRAY.
startTime = time.time()
dictionary = {} # create a dictionary
dictsize = 0
for i in range(-255, 256):
dictionary[str(i)] = dictsize
dictsize += 1
channelsN = img.shape[2]
outputlist = []
outputBytes = bytearray()
listSize = 0
for y in range(img.shape[0]):
for x in range(img.shape[1]):
for c in range(channelsN):
outputlist.append(img[y,x,c])
listSize += 1
print listSize, sys.getsizeof(outputlist)
lastStr = ''
current = outputlist[0]
lastStr = str(int(current))
for i in range(1, listSize):
if dictsize >= 65536:
dictionary = {} # create a dictionary
dictsize = 0
for i in range(-255, 256):
dictionary[str(i)] = dictsize
dictsize += 1
current = outputlist[i]
entry = lastStr+","+str(int(current))
if entry in dictionary:
lastStr = entry
else:
outBits = dictionary[lastStr]
outLow = outBits & 0xFF
outHigh = (outBits >> 8) & 0xFF
outputBytes.append(outLow)
outputBytes.append(outHigh)
dictionary[entry] = dictsize
dictsize += 1
lastStr = str(int(current))
if lastStr in dictionary:
outBits = dictionary[lastStr]
outLow = outBits & 0xFF
outHigh = (outBits >> 8) & 0xFF
outputBytes.append(outLow)
outputBytes.append(outHigh)
endTime = time.time()
# Output the bytes
#
# Include the 'headerText' to identify the type of file. Include
# the rows, columns, channels so that the image shape can be
# reconstructed.
outputFile.write( '%s\n' % headerText )
outputFile.write( '%d %d %d\n' % (img.shape[0], img.shape[1], img.shape[2]) )
outputFile.write( outputBytes )
inSize = img.shape[0] * img.shape[1] * img.shape[2]
outSize = len(outputBytes)
sys.stderr.write( 'Input size: %d bytes\n' % inSize )
sys.stderr.write( 'Output size: %d bytes\n' % outSize )
sys.stderr.write( 'Compression factor: %.2f\n' % (inSize/float(outSize)) )
sys.stderr.write( 'Compression time: %.2f seconds\n' % (endTime - startTime) )