lab3

#https://medium.com/hackernoon/histogram-equalization-in-python-from-scratch-ebb9c8aa3f23

import numpy as np

import matplotlib.pyplot as plt

from PIL import Image

def hist_arr(img_flat, bins):

    histogram = np.zeros(bins)

    for i in img_flat:

        histogram[i] += 1

    return histogram

def cumsum(a):

    a = iter(a)

    b = [next(a)]

    for i in a:

        b.append(b[-1] + i)

    return np.array(b)

def hist_equ_11812418(input_image):

    # Insert code here

    img = Image.open(input_image)

    img_arr = np.asarray(img)

    img_flat = img_arr.flatten() #flatten image to 1D array

    #hist

    # plt.hist(img_flat, bins=50)

    # plt.show()

    input_img_hist = hist_arr(img_flat, 256)

    #print(hist)

    #sum

    # cumulative sum

    # execute the fn

    cumulative_sum = cumsum(input_img_hist)

    # display the cs

    # plt.plot(cs)

    # plt.show()

    #normalize

    cs_norm = ((cumulative_sum - cumulative_sum.min()) * 255)/(cumulative_sum.max() - cumulative_sum.min())

    cs_norm = cs_norm.astype('uint8')

    # plt.plot(cs_norm)

    # plt.show()

    img_new = cs_norm[img_flat]

    #print(img_new)

    img_new_arr = np.reshape(img_new, img_arr.shape)

    output_img_hist = hist_arr(img_new_arr, 256)

    #image file

    ouput_image = Image.fromarray(img_new_arr.astype(np.uint8))

    #hist file

    # return (ouput_image,input_img_hist,output_img_hist)

    return (ouput_image,img_flat,img_new)

 #return (output_image, output_hist, input_hist)

(oimg,input_hist,output_hist) = hist_equ_11812418("Q3_1_2.tif")

oimg.save("output/Q3_1_1_M.tif")

plt.hist(input_hist, bins=256)

plt.savefig('output/input_hist.png')

plt.clf()

plt.hist(output_hist, bins=256)

plt.savefig('output/output_hist.png')

import time

import matplotlib.pyplot as plt

import numpy as np

from PIL import Image

# define hist

hist = np.concatenate((np.linspace(0,10, 8 - 0),

                       np.linspace(10, 0.75, 16 - 8),

                       np.linspace(0.75, 0, 184 - 16),

                       np.linspace(0, 0.5, 200 - 184),

                       np.linspace(0.5, 0, 256 - 200)), axis=0)

spec_hist = np.zeros(256)

def hist_accumulation(hist, bins):

    histogram_in_data_normalized = hist / np.sum(hist)

    # historygram_accumulation

    histogram_in_data_accumulated = np.round(np.add.accumulate(histogram_in_data_normalized) * bins)

    return histogram_in_data_accumulated

def hist_normalization(hist, bins):

    histogram_in_data_normalized = hist / np.sum(hist)

    # historygram_accumulation

    return histogram_in_data_normalized

def hist_equ(img_flat, bins):

    bins_arr = np.arange(bins + 1)

    # generate histogram

    histogram_in_data, histogram_in_index = np.histogram(img_flat, bins=bins_arr)

    # history equ

    # histogram Normalized

    histogram_in_data_normalized = histogram_in_data / np.sum(histogram_in_data)

    # historygram_accumulation

    histogram_in_data_accumulated = np.round(np.add.accumulate(histogram_in_data_normalized) * bins)

    img_after_hist_equ = histogram_in_data_accumulated[img_flat]

    return histogram_in_data_accumulated, histogram_in_data, img_after_hist_equ

def hist_match(hist_in, hist_desired):

    hist_len = len(hist_in)

    hist_out = np.zeros(hist_len)

    def hist_match_index(i):

        index = min(range(hist_len), key=lambda j: abs(hist_desired[i] - hist_in[j]))

        hist_out[i] = index

    # def hist_match_index_par(i):

    #     return min(range(hist_len), key=lambda j: abs(hist_in[j] - hist_desired[i]))

    #

    # c = Parallel( n_jobs = 4 )( delayed( hist_match_index_par )( item ) for item in range(hist_len))

    # return c

    # init pool consume more time

    for i in range(hist_len):

        hist_match_index(i)

    return hist_out

def hist_match_11812418(input_image, spec_hist):

    # Insert code here

    img = Image.open(input_image)

    img_arr = np.asarray(img)

    img_flat = img_arr.flatten()  # flatten image to 1D array

    bins = 256

    # input img hist equ

    input_img_hist_after_hist_equ, input_hist, input_img_after_hist_equ = hist_equ(img_flat, bins)

    # desired_hist_normalized = hist_normalization(spec_hist, bins)

    desired_hist_accumulated = hist_accumulation(spec_hist, bins)

    # in: input_img_hist_after_hist_equ

    # desired: desired_hist_accumulated

    hist_lut = np.array(hist_match(input_img_hist_after_hist_equ, desired_hist_accumulated))

    plt.plot(input_img_hist_after_hist_equ)

    plt.show()

    plt.plot(desired_hist_accumulated)

    plt.show()

    img_after_hist_match = hist_lut[img_flat.astype(int) - 1]

    img_new_arr = np.reshape(input_img_after_hist_equ, img_arr.shape)

    output_hist, histogram_out_index = np.histogram(img_after_hist_match, bins=np.arange(bins + 1))

    return (img_new_arr, output_hist, input_hist)

def plot_hist(hist, filename, title):

    plt.figure(figsize=(8, 6), dpi=300)

    plt.plot(hist)

    plt.title(title)

    plt.savefig(filename)

    plt.show()

if __name__ == '__main__':

    # Q3_2

    start_time = time.time()

    (out_img2, hist_out, hist_in) = hist_match_11812418("Q3_2.tif", hist)

    op_image = Image.fromarray(out_img2.astype(np.uint8))

    print("--- %s seconds ---" % (time.time() - start_time))

    op_image.save("output/img/hist_matching/Q3_2_M.tif")

    op_image.save("output/img/hist_matching/Q3_2_M.png")

    plot_hist(hist_in, "output/img/hist_matching/Q3_2_hist.png", "Histogram before histogram matching")

    plot_hist(hist_out, "output/img/hist_matching/Q3_2_M_hist.png", "Histogram after histogram matching")

    plot_hist(hist, "output/img/hist_matching/hist.png", "Histogram used to match")

import numpy as np

import matplotlib.pyplot as plt

from PIL import Image

import time

def hist_equ(img_flat,bins):

    bins_arr = np.arange(bins + 1)

    # generate histogram

    histogram_in_data, histogram_in_index = np.histogram(img_flat, bins=bins_arr)

    # print(histogram_in_data)

    # history equ

    # histogram Normalized

    histogram_in_data_normalized = histogram_in_data / np.sum(histogram_in_data)

    # historygram_accumulation

    histogram_in_data_accumulated = np.round(np.add.accumulate(histogram_in_data_normalized) * bins)

    # plt.plot(histogram_in_data_accumulated)

    # plt.show()

    img_after_hist_equ = histogram_in_data_accumulated[img_flat]

    return histogram_in_data_accumulated, img_after_hist_equ

def local_hist_equ_11812418(input_image):

    # Insert code here

    img = Image.open(input_image)

    img_arr = np.asarray(img)

    img_flat = img_arr.flatten() #flatten image to 1D array

    bins = 256

    input_img_hist_after_hist_equ, input_img_after_hist_equ = hist_equ(img_flat, bins)

    plt.hist(input_img_after_hist_equ,256)

    plt.show()

    print(input_img_after_hist_equ)

    # #print for test

    # img_new_arr = np.reshape(input_img_after_hist_equ, img_arr.shape)

    # ouput_image = Image.fromarray(img_new_arr.astype(np.uint8))

    # ouput_image.save("output/0311-1.tif")

    return ()

def local_hist_equ(in_img, m_size):

    """

    Implement the local histogram equalization to the input images Q3_3.tif

    """

    img = Image.open(in_img)

    in_img = np.asarray(img)

    row, col = in_img.shape

    out_img = np.zeros(in_img.shape, int)

    bin_num = 256

    bins = range(bin_num + 1)

    for i in list(range(0, row - m_size, m_size)) + [row - m_size - 1]:

        # i = min(i, row - m_size)

        for j in list(range(0, col - m_size, m_size)) + [col - m_size - 1]:

            # j = min(j, col - m_size)

            local_img = in_img[i:i + m_size, j:j + m_size]

            local_hist, _ = np.histogram(local_img.flat, bins=bins, density=True)

            s = np.array([(bin_num - 1) * np.sum(local_hist[:k + 1]) for k in range(bin_num)])

            out_img[i:i + m_size, j:j + m_size] = np.array([s[r] for r in local_img], int).reshape((m_size, m_size))

    return out_img

# local_hist_equ_11812418("Q3_3.tif")

start_time = time.time()

out_img2 = local_hist_equ("Q3_3.tif",2)

ouput_image = Image.fromarray(out_img2.astype(np.uint8))

print("--- %s seconds ---" % (time.time() - start_time))

ouput_image.save("output/0311-2.tif")

 No newline at end of file