How to use windowSize method in fMBT

Best Python code snippet using fMBT_python

main.py

Source:main.py

1# python 3.9.52from math import ceil3import cv24import numpy as np5from functions import *6# define quantization tables7QTY = np.array([[16, 11, 10, 16, 24, 40, 51, 61],  # luminance quantization table8                [12, 12, 14, 19, 26, 48, 60, 55],9                [14, 13, 16, 24, 40, 57, 69, 56],10                [14, 17, 22, 29, 51, 87, 80, 62],11                [18, 22, 37, 56, 68, 109, 103, 77],12                [24, 35, 55, 64, 81, 104, 113, 92],13                [49, 64, 78, 87, 103, 121, 120, 101],14                [72, 92, 95, 98, 112, 100, 103, 99]])15QTC = np.array([[17, 18, 24, 47, 99, 99, 99, 99],  # chrominance quantization table16                [18, 21, 26, 66, 99, 99, 99, 99],17                [24, 26, 56, 99, 99, 99, 99, 99],18                [47, 66, 99, 99, 99, 99, 99, 99],19                [99, 99, 99, 99, 99, 99, 99, 99],20                [99, 99, 99, 99, 99, 99, 99, 99],21                [99, 99, 99, 99, 99, 99, 99, 99],22                [99, 99, 99, 99, 99, 99, 99, 99]])23# define window size24windowSize = len(QTY)25# read image26imgOriginal = cv2.imread('marbles.bmp', cv2.IMREAD_COLOR)27# convert BGR to YCrCb28img = cv2.cvtColor(imgOriginal, cv2.COLOR_BGR2YCR_CB)29width = len(img[0])30height = len(img)31y = np.zeros((height, width), np.float32) + img[:, :, 0]32cr = np.zeros((height, width), np.float32) + img[:, :, 1]33cb = np.zeros((height, width), np.float32) + img[:, :, 2]34# size of the image in bits before compression35totalNumberOfBitsWithoutCompression = len(y) * len(y[0]) * 8 + len(cb) * len(cb[0]) * 8 + len(cr) * len(cr[0]) * 836# channel values should be normalized, hence subtract 12837y = y - 12838cr = cr - 12839cb = cb - 12840# 4: 2: 2 subsampling is used # another subsampling scheme can be used41# thus chrominance channels should be sub-sampled42# define subsampling factors in both horizontal and vertical directions43SSH, SSV = 2, 244# filter the chrominance channels using a 2x2 averaging filter # another type of filter can be used45crf = cv2.boxFilter(cr, ddepth=-1, ksize=(2, 2))46cbf = cv2.boxFilter(cb, ddepth=-1, ksize=(2, 2))47crSub = crf[::SSV, ::SSH]48cbSub = cbf[::SSV, ::SSH]49# check if padding is needed,50# if yes define empty arrays to pad each channel DCT with zeros if necessary51yWidth, yLength = ceil(len(y[0]) / windowSize) * windowSize, ceil(len(y) / windowSize) * windowSize52if (len(y[0]) % windowSize == 0) and (len(y) % windowSize == 0):53    yPadded = y.copy()54else:55    yPadded = np.zeros((yLength, yWidth))56    for i in range(len(y)):57        for j in range(len(y[0])):58            yPadded[i, j] += y[i, j]59# chrominance channels have the same dimensions, meaning both can be padded in one loop60cWidth, cLength = ceil(len(cbSub[0]) / windowSize) * windowSize, ceil(len(cbSub) / windowSize) * windowSize61if (len(cbSub[0]) % windowSize == 0) and (len(cbSub) % windowSize == 0):62    crPadded = crSub.copy()63    cbPadded = cbSub.copy()64# since chrominance channels have the same dimensions, one loop is enough65else:66    crPadded = np.zeros((cLength, cWidth))67    cbPadded = np.zeros((cLength, cWidth))68    for i in range(len(crSub)):69        for j in range(len(crSub[0])):70            crPadded[i, j] += crSub[i, j]71            cbPadded[i, j] += cbSub[i, j]72# get DCT of each channel73# define three empty matrices74yDct, crDct, cbDct = np.zeros((yLength, yWidth)), np.zeros((cLength, cWidth)), np.zeros((cLength, cWidth))75# number of iteration on x axis and y axis to calculate the luminance cosine transform values76hBlocksForY = int(len(yDct[0]) / windowSize)  # number of blocks in the horizontal direction for luminance77vBlocksForY = int(len(yDct) / windowSize)  # number of blocks in the vertical direction for luminance78# number of iteration on x axis and y axis to calculate the chrominance channels cosine transforms values79hBlocksForC = int(len(crDct[0]) / windowSize)  # number of blocks in the horizontal direction for chrominance80vBlocksForC = int(len(crDct) / windowSize)  # number of blocks in the vertical direction for chrominance81# define 3 empty matrices to store the quantized values82yq, crq, cbq = np.zeros((yLength, yWidth)), np.zeros((cLength, cWidth)), np.zeros((cLength, cWidth))83# and another 3 for the zigzags84yZigzag = np.zeros(((vBlocksForY * hBlocksForY), windowSize * windowSize))85crZigzag = np.zeros(((vBlocksForC * hBlocksForC), windowSize * windowSize))86cbZigzag = np.zeros(((vBlocksForC * hBlocksForC), windowSize * windowSize))87for i in range(vBlocksForY):88    for j in range(hBlocksForY):89        yDct[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] = cv2.dct(90            yPadded[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize])91        yq[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] = np.ceil(92            yDct[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] / QTY)93        yZigzag[i * j] += zigzag(94            yq[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize])95yZigzag = yZigzag.astype(np.int16)96# either crq or cbq can be used to compute the number of blocks97for i in range(vBlocksForC):98    for j in range(hBlocksForC):99        crDct[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] = cv2.dct(100            crPadded[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize])101        crq[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] = np.ceil(102            crDct[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] / QTC)103        crZigzag[i * j] += zigzag(104            crq[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize])105        cbDct[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] = cv2.dct(106            cbPadded[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize])107        cbq[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] = np.ceil(108            cbDct[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize] / QTC)109        cbZigzag[i * j] += zigzag(110            cbq[i * windowSize: i * windowSize + windowSize, j * windowSize: j * windowSize + windowSize])111crZigzag = crZigzag.astype(np.int16)112cbZigzag = cbZigzag.astype(np.int16)113# find the run length encoding for each channel114# then get the frequency of each component in order to form a Huffman dictionary115yEncoded = run_length_encoding(yZigzag)116yFrequencyTable = get_freq_dict(yEncoded)117yHuffman = find_huffman(yFrequencyTable)118crEncoded = run_length_encoding(crZigzag)119crFrequencyTable = get_freq_dict(crEncoded)120crHuffman = find_huffman(crFrequencyTable)121cbEncoded = run_length_encoding(cbZigzag)122cbFrequencyTable = get_freq_dict(cbEncoded)123cbHuffman = find_huffman(cbFrequencyTable)124# calculate the number of bits to transmit for each channel125# and write them to an output file126file = open("CompressedImage.asfh", "w")127yBitsToTransmit = str()128for value in yEncoded:129    yBitsToTransmit += yHuffman[value]130crBitsToTransmit = str()131for value in crEncoded:132    crBitsToTransmit += crHuffman[value]133cbBitsToTransmit = str()134for value in cbEncoded:135    cbBitsToTransmit += cbHuffman[value]136if file.writable():137    file.write(yBitsToTransmit + "\n" + crBitsToTransmit + "\n" + cbBitsToTransmit)138file.close()139totalNumberOfBitsAfterCompression = len(yBitsToTransmit) + len(crBitsToTransmit) + len(cbBitsToTransmit)140print(141    "Compression Ratio is " + str(...

LBPfuncs.py

Source:LBPfuncs.py

1import cv22import numpy as np34import get_LBP_from_Image as LBP567def region_flag(Sub, yt, xt, windowSize, region_thresh):8    return np.sum(Sub[yt:yt + windowSize, xt:xt + windowSize]) > (windowSize ** 2) / region_thresh910def region_flag_out(out, yt, xt, windowSize, region_thresh_out):11    return np.sum(out[yt:yt + windowSize, xt:xt + windowSize]) <= region_thresh_out12131415def MyResize(I, factor):16    w = I.shape[0]17    h = I.shape[1]18    w0 = int(w / factor)19    h0 = int(h / factor)20    out = cv2.resize(I, (w0, h0), interpolation=cv2.INTER_LINEAR)21    return out222324def calc_regional_LBP(target, background, lbp, out, yt, xt, windowSize):25    t_windowed = target[yt:yt + windowSize, xt:xt + windowSize]26    b_windowed = background[yt:yt + windowSize, xt:xt + windowSize]27    t_LBP_map = lbp.lbp_uniform(t_windowed) # LBPçä»·æ¨¡å¼ç¹å¾28    b_LBP_map = lbp.lbp_uniform(b_windowed)29    t_hist = lbp.get_uniform_hist(t_LBP_map) # è·åLBPç¹å¾å½ä¸åç»è®¡ç´æ¹å¾30    b_hist = lbp.get_uniform_hist(b_LBP_map)31    d = LBP.chi2_distance(t_hist, b_hist)32    # print((j, i))33    out[yt:yt + windowSize, xt:xt + windowSize] += np.ones((windowSize, windowSize)) * d343536def CompareLBP(target, background, Sub, windowSize=32, step=4, region_thresh=3, decay=0.05):37    [h, w] = [target.shape[0], target.shape[1]]38    x_iter = int((w - windowSize) / step) + 139    y_iter = int((h - windowSize) / step) + 140    x_left = w - windowSize - step * (x_iter - 1)41    y_left = h - windowSize - step * (y_iter - 1)42    out = np.zeros((h, w))43    lbp = LBP.LBP()44    flag = 045    deviation = decay46    deviation1 = deviation * 0.7547    deviation2 = deviation * 0.7548    deviation3 = deviation * 0.7549    deviation4 = deviation * 0.7550    for j in range(y_iter):51        for i in range(x_iter):52            xt = i * step53            yt = j * step54            CalcFlag = (np.sum(Sub[yt:yt + windowSize, xt:xt + windowSize]) > (windowSize ** 2) / region_thresh)55            if j < y_iter - 2 and j > 1 and i < x_iter - 2 and i > 1:56                CalcFlag = CalcFlag or region_flag(Sub, yt + windowSize, xt - windowSize, windowSize, region_thresh)57                CalcFlag = CalcFlag or region_flag(Sub, yt + windowSize, xt, windowSize, region_thresh)58                CalcFlag = CalcFlag or region_flag(Sub, yt + windowSize, xt + windowSize, windowSize, region_thresh)59                CalcFlag = CalcFlag or region_flag(Sub, yt, xt + windowSize, windowSize, region_thresh)60                CalcFlag = CalcFlag and region_flag_out(out, yt + windowSize, xt - windowSize, windowSize, 1)61                CalcFlag = CalcFlag and region_flag_out(out, yt + windowSize, xt, windowSize, 1)62                CalcFlag = CalcFlag and region_flag_out(out, yt + windowSize, xt + windowSize, windowSize, 1)63                CalcFlag = CalcFlag and region_flag_out(out, yt, xt + windowSize, windowSize, 1)64            if CalcFlag:65                flag += 366                if flag > 0:67                    calc_regional_LBP(target, background, lbp, out, yt, xt, windowSize)68                    69                    if j < y_iter - 2 and j > 1 and i < x_iter - 2 and i > 1:70                        # if not region_flag(Sub, yt - windowSize, xt - windowSize, windowSize, region_thresh):71                        if region_flag(out, yt - windowSize, xt - windowSize, windowSize, region_thresh * deviation1):72                            calc_regional_LBP(target, background, lbp, out, yt - windowSize, xt - windowSize,73                                              windowSize)74                        if region_flag(out, yt - windowSize, xt, windowSize, region_thresh * deviation2):75                            calc_regional_LBP(target, background, lbp, out, yt - windowSize, xt, windowSize)76                        if region_flag(out, yt - windowSize, xt + windowSize, windowSize, region_thresh * deviation3):77                            calc_regional_LBP(target, background, lbp, out, yt - windowSize, xt + windowSize,78                                              windowSize)79                        if region_flag(out, yt, xt - windowSize, windowSize, region_thresh * deviation4):80                            calc_regional_LBP(target, background, lbp, out, yt, xt - windowSize, windowSize)81                        if region_flag(out, yt, xt + windowSize, windowSize, region_thresh * deviation):82                            calc_regional_LBP(target, background, lbp, out, yt, xt + windowSize, windowSize)83                        if region_flag(out, yt + windowSize, xt - windowSize, windowSize, region_thresh * deviation):84                            calc_regional_LBP(target, background, lbp, out, yt + windowSize, xt - windowSize,85                                              windowSize)86                        if region_flag(out, yt + windowSize, xt, windowSize, region_thresh * deviation):87                            calc_regional_LBP(target, background, lbp, out, yt + windowSize, xt, windowSize)88                        if region_flag(out, yt + windowSize, xt + windowSize, windowSize, region_thresh * deviation):89                            calc_regional_LBP(target, background, lbp, out, yt + windowSize, xt + windowSize,90                                              windowSize)9192                    flag -= 193                # LBP.ShowSubIm("out", out, out[yt:yt + windowSize, xt:xt + windowSize])949596    if x_left:97        for j in range(y_iter):98            yt = j * step99            t_windowed = target[yt:yt + windowSize, w - x_left:w]100            b_windowed = background[yt:yt + windowSize, w - x_left:w]101            t_LBP_map = lbp.lbp_revolve(t_windowed) # LBPæè½¬ä¸åæ¨¡å¼ç¹å¾102            b_LBP_map = lbp.lbp_revolve(b_windowed)103            t_hist = lbp.get_revolve_hist(t_LBP_map) # LBPæè½¬ä¸åå½ä¸åç»è®¡ç´æ¹å¾104            b_hist = lbp.get_revolve_hist(b_LBP_map)105            d = LBP.chi2_distance(t_hist, b_hist)106            out[yt:yt + windowSize, w - x_left:w] += np.ones((windowSize, x_left)) * d107    if y_left:108        for i in range(x_iter):109            xt = i * step110            t_windowed = target[h - y_left:h, xt:xt + windowSize]111            b_windowed = background[h - y_left:h, xt:xt + windowSize]112            t_LBP_map = lbp.lbp_revolve(t_windowed)113            b_LBP_map = lbp.lbp_revolve(b_windowed)114            t_hist = lbp.get_revolve_hist(t_LBP_map)115            b_hist = lbp.get_revolve_hist(b_LBP_map)116            d = LBP.chi2_distance(t_hist, b_hist)117            out[h - y_left:h, xt:xt + windowSize] += np.ones((y_left, windowSize)) * d118    if y_left and x_left:119        t_windowed = target[h - y_left:h, w - x_left:w]120        b_windowed = background[h - y_left:h, w - x_left:w]121        t_LBP_map = lbp.lbp_revolve(t_windowed)122        b_LBP_map = lbp.lbp_revolve(b_windowed)123        t_hist = lbp.get_revolve_hist(t_LBP_map)124        b_hist = lbp.get_revolve_hist(b_LBP_map)125        d = LBP.chi2_distance(t_hist, b_hist)126        out[h - y_left:h, w - x_left:w] += np.ones((y_left, x_left)) * d127    # out = out / np.max(out)
...

match_features.py

Source:match_features.py

1import cv22import numpy as np3def match_features(feature_coords1, feature_coords2, image1, image2, windowSize = None):4    if windowSize == None :5        windowSize = 206        7        dictFeature2To1 = {}8        dictFeature1To2 = {}9        image1 = cv2.cvtColor(image1, cv2.COLOR_RGB2GRAY)10        image2 = cv2.cvtColor(image2, cv2.COLOR_RGB2GRAY)11        for i in feature_coords1:12            windowIn1 = image1[i[0] - windowSize:i[0] + windowSize, i[1] - windowSize:i[1] + windowSize]13            if (windowIn1.shape != (windowSize * 2, windowSize * 2)): continue14            maxVal = -115            index = [0, 0]16            for j in feature_coords2:17                windowIn2 = image2[j[0] - windowSize:j[0] + windowSize, j[1] - windowSize:j[1] + windowSize]18                if (windowIn2.shape != (windowSize * 2, windowSize * 2)): continue19                product = np.mean((windowIn1 - windowIn1.mean()) * (windowIn2 - windowIn2.mean()))20                stds = windowIn1.std() * windowIn2.std()21                product /= stds22                if maxVal < product:23                    maxVal = product24                    index = j25            dictFeature2To1[i] = index26        for i in feature_coords2:27            windowIn2 = image2[i[0] - windowSize:i[0] + windowSize, i[1] - windowSize:i[1] + windowSize]28            if (windowIn2.shape != (windowSize * 2, windowSize * 2)): continue29            maxVal = -130            index = [0, 0]31            for j in feature_coords1:32                windowIn1 = image1[j[0] - windowSize:j[0] + windowSize, j[1] - windowSize:j[1] + windowSize]33                if (windowIn1.shape != (windowSize * 2, windowSize * 2)): continue34                product = np.mean((windowIn2 - windowIn2.mean()) * (windowIn1 - windowIn1.mean()))35                stds = windowIn2.std() * windowIn1.std()36                product /= stds37                if maxVal < product:38                    maxVal = product39                    index = j40            dictFeature1To2[i] = index41    else:42        windowSize = windowSize43        dictFeature2To1 = {}44        dictFeature1To2 = {}45        image1 = cv2.cvtColor(image1, cv2.COLOR_RGB2GRAY)46        image2 = cv2.cvtColor(image2, cv2.COLOR_RGB2GRAY)47        for i in feature_coords1:48            windowIn1 = image1[i[0] - windowSize:i[0] + windowSize, i[1] - windowSize:i[1] + windowSize]49            if (windowIn1.shape != (windowSize * 2, windowSize * 2)): continue50            maxVal = -151            index = [0,0]52            for j in feature_coords2:53                windowIn2 = image2[j[0] - windowSize:j[0] + windowSize, j[1] - windowSize:j[1] + windowSize]54                if (windowIn2.shape != (windowSize * 2, windowSize * 2)): continue55                product = np.mean((windowIn1 - windowIn1.mean()) * (windowIn2 - windowIn2.mean()))56                stds = windowIn1.std() * windowIn2.std()57                product /= stds58                if maxVal < product:59                    maxVal = product60                    index = j61            dictFeature2To1[i] = index62        for i in feature_coords2:63            windowIn2 = image2[i[0] - windowSize:i[0] + windowSize, i[1] - windowSize:i[1] + windowSize]64            if (windowIn2.shape != (windowSize * 2, windowSize * 2)):continue65            maxVal = -166            index = [0,0]67            for j in feature_coords1:68                windowIn1 = image1[j[0] - windowSize:j[0] + windowSize, j[1] - windowSize:j[1] + windowSize]69                if (windowIn1.shape != (windowSize * 2, windowSize * 2)):continue70                product = np.mean((windowIn2 - windowIn2.mean()) * (windowIn1 - windowIn1.mean()))71                stds = windowIn2.std() * windowIn1.std()72                product /= stds73                if maxVal < product:74                    maxVal = product75                    index = j76            dictFeature1To2[i] = index77    matches = list()78    listInPairs = list()79    for i in dictFeature2To1.keys():80        temp = dictFeature2To1[i]81        if dictFeature1To2[temp] == i:82            index1 = feature_coords1.index(i)83            index2 = feature_coords2.index(temp)84            listInPairs.append([index1,index2])85            matches.append(i)86            matches.append(temp)87    print len(matches)...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.