Cfrgtkky

I am trying to compress the images using the K-means algorithm but the size of some .jpg images have increased after compression. How can I do it for .jpg and .jpeg images. I have saved the jpg and jpeg images in png format before applying the compression

for f in os.listdir('.'):

    if f.endswith('.png'):

        image = io.imread(f,0)

        rows = image.shape[0]

        cols = image.shape[1]



        pixels = image.reshape(image.shape[0] * image.shape[1], image.shape[2])

        kmeans = MiniBatchKMeans(n_clusters=128, n_init=10, max_iter=200)

        kmeans.fit(pixels)



        clusters = np.asarray(kmeans.cluster_centers_, dtype=np.uint8)

        labels = np.asarray(kmeans.labels_, dtype=np.uint8)

        labels = labels.reshape(rows, cols)

        colored = clusters[labels]



        #  np.save('codebook'+f+'.npy', clusters)

        io.imsave('compressed_' + f, colored)



        img1 = mpimg.imread(f,0)

        img2 = mpimg.imread('compressed_' + f,0)

        fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 10))

        ax1.imshow(img1)

        ax1.set_title('Original image')

        ax2.imshow(img2)

        ax2.set_title('Compressed image')

        plt.show()



        fig, ax = plt.subplots(2, 1)



        img = cv2.imread(f, 0)

        ax[0].hist(img.ravel(), 256, [0, 256]);

        ax[0].set_title("Original image")

        img1 = cv2.imread('compressed_' + f,0)

        ax[1].hist(img1.ravel(), 256, [0, 256]);

        ax[1].set_title("Compressed image")

        plt.show()



        print('size of original image: ', int(os.stat(f).st_size / 1024), 'kB')

        print('size of compressed image:', int(os.stat('compressed_' + f).st_size / 1024), 'kB')

If I understand your approach correctly, you use K-means clustering to reduce the number of colors in the image by finding areas (clusters) of similar colored pixels and grouping them together.

While this can theoretically reduce the filesize of the image, saving it again as a jpeg applies a whole different image compression algorithm on the reduced image, which cannot take full advantage of big areas of the same color with sharp edges. It will necessarily "blur" the image, and in some cases, this can even lead to an increased file size.

Try storing the reduced image in a different format (e.g. as a png file), that can make use of big, evenly colored areas.