Spectral Color Correction

[xeggbee]

Looking for a help, what can else can be done in order to get best results in the output image.

--Took the picture with Iphone X, orignal format is DNG transformed to jpg, colorcheck is in the image.
--using spectral information to build the CMFS based in the data: https://ohlab.kic.ac.jp/index/mobilespec_database

First, detection:

Input image:

COLOUR_CHECKER_IMAGES = [
    colour.cctf_decoding(RGB)
    for path in COLOUR_CHECKER_IMAGE_PATHS
]
for image in COLOUR_CHECKER_IMAGES:
    colour.plotting.plot_image(colour.cctf_encoding(image));


methodo = 'Finlayson 2015' #'Cheung 2004', 'Finlayson 2015', 'Vandermonde'
dfreal = pd.DataFrame()
listareal = []
SWATCHES = []
emptylist = []


for image in COLOUR_CHECKER_IMAGES:
    for colour_checker_swatches_data in detect_colour_checkers_segmentation(
        image, additional_data=True):
        swatch_colours, colour_checker_image, swatch_masks = (
            colour_checker_swatches_data.values)
        SWATCHES.append(swatch_colours)

        # Using the additional data to plot the colour checker and masks.
        masks_i = np.zeros(colour_checker_image.shape)
        for i, mask in enumerate(swatch_masks):
            masks_i[mask[0]:mask[1], mask[2]:mask[3], ...] = 1
        colour.plotting.plot_image(
            colour.cctf_encoding(
                np.clip(colour_checker_image + masks_i * 0.25, 0, 1)))

Correction

1 - capture all specttral distribution of colour checker

if SWATCHES != emptylist:
    sd1 = SDS_COLOURCHECKERS['BabelColor Average']['dark skin']
    sd2 = SDS_COLOURCHECKERS['BabelColor Average']['light skin']
    sd3 = SDS_COLOURCHECKERS['BabelColor Average']['blue sky']
    sd4 = SDS_COLOURCHECKERS['BabelColor Average']['foliage']
    sd5 = SDS_COLOURCHECKERS['BabelColor Average']['blue flower']
    sd6 = SDS_COLOURCHECKERS['BabelColor Average']['bluish green']
    sd7 = SDS_COLOURCHECKERS['BabelColor Average']['orange']
    sd8 = SDS_COLOURCHECKERS['BabelColor Average']['purplish blue']
    sd9 = SDS_COLOURCHECKERS['BabelColor Average']['moderate red']
    sd10 =SDS_COLOURCHECKERS['BabelColor Average']['purple']
    sd11 =SDS_COLOURCHECKERS['BabelColor Average']['yellow green']
    sd12 =SDS_COLOURCHECKERS['BabelColor Average']['orange yellow']
    sd13 =SDS_COLOURCHECKERS['BabelColor Average']['blue']
    sd14 =SDS_COLOURCHECKERS['BabelColor Average']['green']
    sd15 =SDS_COLOURCHECKERS['BabelColor Average']['red']
    sd16 =SDS_COLOURCHECKERS['BabelColor Average']['yellow']
    sd17 =SDS_COLOURCHECKERS['BabelColor Average']['magenta']
    sd18 =SDS_COLOURCHECKERS['BabelColor Average']['cyan']
    sd19 =SDS_COLOURCHECKERS['BabelColor Average']['white 9.5 (.05 D)']
    sd20 =SDS_COLOURCHECKERS['BabelColor Average']['neutral 8 (.23 D)']
    sd21 =SDS_COLOURCHECKERS['BabelColor Average']['neutral 6.5 (.44 D)']
    sd22 =SDS_COLOURCHECKERS['BabelColor Average']['neutral 5 (.70 D)']
    sd23 =SDS_COLOURCHECKERS['BabelColor Average']['neutral 3.5 (1.05 D)']
    sd24 =SDS_COLOURCHECKERS['BabelColor Average']['black 2 (1.5 D)']

2- since I know the Illuminant, collect during the data collection, i tried using different native illumiant and build my own illuminant, same results.

    ############################## Illuminant
    illuminant = colour.SDS_ILLUMINANTS['D50']

    #illuminant = colour.SDS_ILLUMINANTS['FL2']
    
    CCT_D65 = (6038) * 1.4388 / 1.4380
    xy = CCT_to_xy_CIE_D(CCT_D65)
    with numpy_print_options(suppress=True):
        az = colour.sd_CIE_illuminant_D_series(xy)
    illuminant = az
    #print (CCT_D65, " + ", xy)

3-using the spectral sentivities of Iphone to build the cmfs

    ############################## CMFS - iphone x
    ix = pd.read_csv('SpectralSensitivity/iphone/iPhoneX.txt', sep='\t', header=None)
    del ix[3]
    del ix[4]
    del ix[5]

    ix = ix.rename(columns={0: 'R', 1: 'G', 2: 'B'})
    w = colour.SpectralShape(400, 700, 10).range()
    cmfs = colour.MultiSpectralDistributions(ix,w, name="Red")
    ##cmfs = colour.MSDS_CMFS['CIE 1964 10 Degree Standard Observer']
    ##############################

4-followed the Color Correction method

  # Calculating the sample spectral distribution *CIE XYZ* tristimulus values.
    cor1 = colour.sd_to_XYZ(sd1, cmfs, illuminant)
    cor2 = colour.sd_to_XYZ(sd2, cmfs, illuminant)
    cor3 = colour.sd_to_XYZ(sd3, cmfs, illuminant)
    cor4 = colour.sd_to_XYZ(sd4, cmfs, illuminant)
    cor5 = colour.sd_to_XYZ(sd5, cmfs, illuminant)
    cor6 = colour.sd_to_XYZ(sd6, cmfs, illuminant)
    cor7 = colour.sd_to_XYZ(sd7, cmfs, illuminant)
    cor8 = colour.sd_to_XYZ(sd8, cmfs, illuminant)
    cor9 = colour.sd_to_XYZ(sd9, cmfs, illuminant)
    cor10 = colour.sd_to_XYZ(sd10, cmfs, illuminant)
    cor11 = colour.sd_to_XYZ(sd11, cmfs, illuminant)
    cor12 = colour.sd_to_XYZ(sd12, cmfs, illuminant)
    cor13 = colour.sd_to_XYZ(sd13, cmfs, illuminant)
    cor14 = colour.sd_to_XYZ(sd14, cmfs, illuminant)
    cor15 = colour.sd_to_XYZ(sd15, cmfs, illuminant)
    cor16 = colour.sd_to_XYZ(sd16, cmfs, illuminant)
    cor17 = colour.sd_to_XYZ(sd17, cmfs, illuminant)
    cor18 = colour.sd_to_XYZ(sd18, cmfs, illuminant)
    cor19 = colour.sd_to_XYZ(sd19, cmfs, illuminant)
    cor20 = colour.sd_to_XYZ(sd20, cmfs, illuminant)
    cor21 = colour.sd_to_XYZ(sd21, cmfs, illuminant)
    cor22 = colour.sd_to_XYZ(sd22, cmfs, illuminant)
    cor23 = colour.sd_to_XYZ(sd23, cmfs, illuminant)
    cor24 = colour.sd_to_XYZ(sd24, cmfs, illuminant)

    patch =[]
    patch = np.array([cor1,cor2,cor3,cor4,cor5,cor6,cor7,cor8,cor9,cor10,cor11,cor12,cor13,cor14,cor15,cor16,cor17,cor18,cor19,cor20,cor21,cor22,cor23,cor24])
    

    patch = patch/100 **---> already tried differnt values**
    print ("Antes da correcao - delta E")
    reais = XYZ_to_Lab(SWATCHES) 
    reais
    cc = XYZ_to_Lab(patch) 
    cc
    print (delta_E(reais, cc, method='CIE 2000')  )
    print (np.sum(delta_E(reais, cc, method='CIE 2000')))

    
    ##reshape the SWATCHES
    SWATCHES
    arr = np.array(SWATCHES)
    arr.shape
    newarr = arr.reshape(24,3)
    newarr.shape

    ##CCM method
    #colour.characterisation.matrix_colour_correction_Finlayson2015(newarr, REFERENCE_SWATCHES) 
    
    ## to new Image
    newimg = 0
    newimg = colour.colour_correction(RGB, newarr, patch, method = methodo);
    #method (Union[Literal['Cheung 2004', 'Finlayson 2015', 'Vandermonde'], str]) – Computation method.
    ##Plot corrected ImageMode
    CCT_D65 = illuminant
    print ("Cor corrigida - no iluminante:", CCT_D65)
    #colour.plotting.plot_image(newimg)

    ## nome da imagem
    path2 = "/kit_teste4/spectral_correction/novo"
    file_name = os.path.basename(paths)
    print (file_name)
    nomeoriginal = file_name[:-4]
    novo_name = str(nomeoriginal) + str("_2_") + str("_") + str(methodo) + str("_.png")

    #salva iamgem
    write_image_Imageio(newimg, os.path.join(path2, novo_name), "float32")

result:

[KelSolaar]

Hello @xeggbee,

Are you trying to generate a set of reference patches using your iPhone sensitivities and colour correct using those?

Cheers,

Thomas