Skip to content

Fix #23896 Vector Quantization Example increases rather than decrea… #24484

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 5 commits into from
Closed

Fix #23896 Vector Quantization Example increases rather than decrea… #24484

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
3de1934
Possible Fix Vector Quantization Example increases rather than decrea…
sxsmg Sep 20, 2022
43446da
minor fix for failing test
sxsmg Sep 20, 2022
457b2ab
failing test: ran black for formatting
sxsmg Sep 21, 2022
e294adf
minor fix linting error
sxsmg Sep 21, 2022
68a32c0
Merge branch 'main' into main
sxsmg Sep 21, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
74 changes: 70 additions & 4 deletions examples/cluster/plot_face_compress.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,13 +26,13 @@

face = face(gray=True)
except ImportError:
face = sp.face(gray=True)
face = sp.misc.face(gray=True)

n_clusters = 5
N_CLUSTERS = 5
np.random.seed(0)

X = face.reshape((-1, 1)) # We need an (n_sample, n_feature) array
k_means = cluster.KMeans(n_clusters=n_clusters, n_init=4)
k_means = cluster.KMeans(n_clusters=N_CLUSTERS, n_init=4)
k_means.fit(X)
values = k_means.cluster_centers_.squeeze()
labels = k_means.labels_
Expand All @@ -53,7 +53,7 @@
plt.imshow(face_compressed, cmap=plt.cm.gray, vmin=vmin, vmax=vmax)

# equal bins face
regular_values = np.linspace(0, 256, n_clusters + 1)
regular_values = np.linspace(0, 256, N_CLUSTERS + 1)
regular_labels = np.searchsorted(regular_values, face) - 1
regular_values = 0.5 * (regular_values[1:] + regular_values[:-1]) # mean
regular_face = np.choose(regular_labels.ravel(), regular_values, mode="clip")
Expand All @@ -76,3 +76,69 @@
plt.axvline(0.5 * (center_1 + center_2), color="b", linestyle="--")

plt.show()

#Word Of Caution:

#Theoretically, the above example should demonstrate compression.
#But there's a catch when we try to check the size of images;
#the compressed image actually ends up taking more memory
#space than the image itself. Even though the number of
#unique values in the image is greater compared to the
#compressed image.The catch comes from the fact that
#numpy arrays are contiguous and are required to have
#preallocated memory for each object. This is where
#the numpy speed comes from. But for example, as above,
#it has caused problems. The kmeans does reduce the
#dimensionality or number of unique values required to
#represent the image,but they are still stored in the same length nparray.

#Colored Image version With actual Compression:

face = sp.misc.face(gray=False)

N_CLUSTERS = 64 # clusters of colors
np.random.seed(0)

plt.imshow(face)

w, h, d = original_shape = tuple(
face.shape
) # a colored image contains 3 stacked matrices

face = np.array(face, dtype=np.float64) / 255 # small number easier to multiply
face_array = np.reshape(face, (w * h, d)) # reshape the image for Kmeans
kmeans = cluster.KMeans(n_clusters=N_CLUSTERS, random_state=0).fit(face_array)
labels = kmeans.predict(face_array)


plt.figure(1)
plt.clf()
plt.axis("on")
plt.title("Original image")
plt.imshow(face)

plt.figure(2)
plt.clf()
plt.axis("on")
plt.title("Quantised image")
img_re = kmeans.cluster_centers_[labels].reshape(w, h, -1)
plt.imshow(img_re)

plt.figure(3)
plt.clf()
plt.axis("on")
plt.title("Quantised image Compressed")
img_re_compressed = (img_re * 255).astype("uint8")
plt.imshow(img_re_compressed)

# number of elements multiplied by byte size of each element
img_re_size = img_re.itemsize * img_re.size
img_re_c_size = img_re_compressed.itemsize * img_re_compressed.size
# number of unique elements reduce in compressed image due to clustering
face_uniq = len(np.unique(face))
img_re_c_uniq = len(np.unique(img_re_compressed))

print(f"Size of reconstructed image: {img_re_size} bytes")
print(f"Size of compressed reconstructed image: {img_re_c_size} bytes")
print(f"number of unique values in original image: {face_uniq}")
print(f"number of unique values in reconstructed image: {img_re_c_uniq}")