_miam/_miamG/miam/aesthetics/LightnessAesthetics.py

# import
# ------------------------------------------------------------------------------------------
from .. import image
import miam
import copy
import numpy as np
import matplotlib.pyplot as plt
# miam import
import miam.image.Image as MIMG
import miam.processing.ColorSpaceTransform as MCST
import miam.histogram.Histogram  as MHIST
import miam.aesthetics.LightnessAesthetics as MLAC
# ------------------------------------------------------------------------------------------
# MIAM project 2020
# ------------------------------------------------------------------------------------------
# author: remi.cozot@univ-littoral.fr
# ------------------------------------------------------------------------------------------
class LightnessAesthetics(object):
    """description of class"""
    def __init__(self, centroids, **kwargs):
        self.lightnessClasses =  centroids

        self.nbClass = len(self.lightnessClasses)
        self.nbBin = self.lightnessClasses[0].histValue.shape[0]

    def plot(self,ax):
        for j,aestheticsClass in enumerate(self.lightnessClasses):
            sum, sumB= 0.0,0.0
            # compute mean
            for lightnessValue, binValue in enumerate(aestheticsClass.histValue):
                sumB = sumB + binValue
                sum= sum + binValue*aestheticsClass.edgeValue[lightnessValue+1]
            ax.plot(aestheticsClass.edgeValue[1:],aestheticsClass.histValue,miam.pltcolor[j])
            ax.plot(sum/sumB,0,miam.pltcolor[j]+'o')

    def computeImageClass(self, image):
        # if required transform in Lab
        # note lightness class are in Lab
        image = MCST.ColorSpaceTransform().compute(image,dest='Lab')
        # compute lightness histogram
        hist = MHIST.Histogram.build(image,MIMG.channel.channel.L,nbBins=self.nbBin)
        # compute distance with classes
        distance = []
        for i, aestheticsSignature in enumerate(self.lightnessClasses) :
            dist = MHIST.Histogram.computeDistance(aestheticsSignature,hist)
            distance.append(dist)
        return distance

    def projectImage(self,image,nbClass='all'):
        distance = self.computeImageClass(image)
        hLab = MHIST.Histogram.build(MCST.ColorSpaceTransform().compute(image,dest='Lab'),MIMG.channel.channel.L,nbBins=self.nbBin)
        res = copy.deepcopy(hLab)
        res.histValue = np.zeros(res.histValue.shape)
        if not nbClass: nbClass ="all"
        if isinstance(nbClass, str):
            if nbClass == "all":
                for i, aestheticsSignature in enumerate(self.lightnessClasses) :
                    res = MHIST.Histogram.add(res, MHIST.Histogram.scale((1.0 - distance[i]),aestheticsSignature))
            elif nbClass =="first":
                idxSort = np.argsort(np.asarray(distance))
                res = MHIST.Histogram.scale((1.0 - distance[idxSort[0]]), self.lightnessClasses[idxSort[0]])
        if isinstance(nbClass,range):
            idxSort = np.argsort(np.asarray(distance))
            for i in nbClass:
                res = MHIST.Histogram.add(res,MHIST.Histogram.scale((1.0 - distance[idxSort[i]]), self.lightnessClasses[idxSort[i]]))
        return res.normalise(norm='dot')

    def readLightnessClasses(fileName, className =  None):
        """ form np.load recover histogram of lightness classes"""
        if not className: className=['black','shadow','medium','highlight','white']

        # from np.histogram to Histograms class
        centroids = np.load(fileName)

        nbClass, nbBin =centroids.shape
        edges = np.linspace(0,100,nbBin+1)

        # compute min to set name
        mean= []
        for j,c in enumerate(centroids):
            sum, sumB= 0.0,0.0
            # compute mean
            for i, binValue in enumerate(c):
                sumB = sumB + binValue
                sum= sum + binValue*edges[i+1]
            mean.append(sum/sumB)
        
        # sortIndex mean
        sortedIndex = np.argsort(mean)

        # build Histogram object
        lightnessClasses = []
        for i, j in enumerate(sortedIndex):
            name = className[i]
            histValue = centroids[j]
            edgeValue = copy.deepcopy(edges)
            channel = MIMG.channel.channel.L
            colorSpace = MIMG.ColorSpace.ColorSpace.buildLab()
            log       = False

            # build a histogram
            lightnessClasses.append(MHIST.Histogram(histValue, edgeValue, name,channel,logSpace = log))

        return LightnessAesthetics(lightnessClasses)