class test_DatasetUtils(unittest.TestCase):
    def setUp(self) -> None:
        super().setUp()
        self.datasetUtils = DataUtils()

    def test_generateCFAImages(self):
        rgbImages = self.datasetUtils.loadKodakDataset()
        cfaImages, image_size = self.datasetUtils.convertDatasetToCFA(
            rgbImages)
        self.assertIsNotNone(cfaImages)

    def test_saveNikonImage(self):
        self.datasetUtils.saveImageNikon()
Ejemplo n.º 2
0
class TestORCTOptimization(unittest.TestCase):
    def __init__(self, methodName: str = ...) -> None:
        super().__init__(methodName)
        self.datasetUtils = DataUtils()
        self.evaluation = Evaluation()

    def opt_func(self, X, twoComplement):
        n_particles = X.shape[0]  # number of particles
        costs = []
        for x in X:
            x = np.reshape(x, [2, 2])
            filtered = compute_orct2(compute_orct1(twoComplement, x), x)
            filtered = (filtered + 255) / 2

            def inverseFunction(data):
                data = data.astype('float32') * 2 - 255
                data = compute_orct2inverse(data, x)
                data = compute_orct1inverse(data, x)
                return data

            sampleFunctionReverse = inverseFunction
            psnr, ssim, jpeg2000CompressionRatioAfter, jpeg2000CompressionRatioBefore = self.evaluation.evaluate(
                filtered, twoComplement, sampleFunctionReverse)
            cost = np.abs(
                (1 / psnr) * (1 / ssim) * 1 / (jpeg2000CompressionRatioAfter))
            costs.append(cost)
        return np.array(costs)

    def test_orct12_optimization(self):
        bayer = self.datasetUtils.readCFAImages()

        twoComplement = self.datasetUtils.twoComplementMatrix(bayer)
        twoComplement = twoComplement.astype("float32")

        options = {'c1': 0.5, 'c2': 0.1, 'w': 0.9}
        optimizer = GlobalBestPSO(n_particles=10,
                                  dimensions=4,
                                  options=options)

        costFunction = partial(self.opt_func, twoComplement=twoComplement)
        cost, pos = optimizer.optimize(costFunction, iters=30)

        pass

    def test_orct12_(self):
        pos = np.asarray([0.39182592, 0.23747258, 0.51497874, -0.08751142])
        x = np.reshape(pos, [2, 2])
        # x = np.asarray([[0.05011018, -0.53709484],
        #                 [-1.1104253, -0.30699651]])
        bayer = self.datasetUtils.readCFAImages()

        twoComplement = self.datasetUtils.twoComplementMatrix(bayer)
        twoComplement = twoComplement.astype("float32")

        filtered = compute_orct2(compute_orct1(twoComplement, x), x)

        filtered = (filtered + 255) / 2

        def inverseFunction(data):
            data = data.astype('float32') * 2 - 255
            data = compute_orct2inverse(data, x)
            data = compute_orct1inverse(data, x)
            return data

        sampleFunctionReverse = inverseFunction
        self.evaluation.evaluate(filtered, twoComplement,
                                 sampleFunctionReverse)
        pass

    def test_ocrtOptimizedWithDataset(self):
        pos = np.asarray([0.39182592, 0.23747258, 0.51497874, -0.08751142])
        x = np.reshape(pos, [2, 2])
        rgbImages = self.datasetUtils.loadKodakDataset()
        cfaImages, image_size = self.datasetUtils.convertDatasetToCFA(
            rgbImages)
        bayer = cfaImages[2, :, :]

        twoComplement = self.datasetUtils.twoComplementMatrix(bayer)
        twoComplement = twoComplement.astype("float32")

        filtered = compute_orct2plus3(compute_orct1(twoComplement, x), x)

        filtered = (filtered + 255) / 2

        def inverseFunction(data):
            data = data.astype('float32') * 2 - 255
            data = compute_orct2plus3inverse(data, x)
            data = compute_orct1inverse(data, x)
            return data

        sampleFunctionReverse = inverseFunction
        self.evaluation.evaluate(filtered, bayer, sampleFunctionReverse)
        pass

    def test_multiObjOptimization(self):
        algorithm = NSGA2(pop_size=10,
                          n_offsprings=10,
                          sampling=get_sampling("real_random"),
                          crossover=get_crossover("real_sbx", prob=0.9,
                                                  eta=15),
                          mutation=get_mutation("real_pm", eta=20),
                          eliminate_duplicates=True)
        termination = get_termination("n_gen", 5)
        bayer = self.datasetUtils.readCFAImages()
        twoComplement = self.datasetUtils.twoComplementMatrix(bayer)
        twoComplement = twoComplement.astype("float32")
        problem = MyProblem(twoComplement)
        res = minimize(problem,
                       algorithm,
                       termination,
                       save_history=True,
                       verbose=True)

        # Objective Space
        res.F = 1 / res.F
        plot = Scatter(title="Objective Space")
        plot.add(res.F)
        plot.show()
        print("Best filter{}".format(np.reshape(res.opt[-1].X, [2, 2])))