58a220cb0fa5ea7d86486f1b3af8036b905de081,cellprofiler/modules/applythreshold.py,ApplyThreshold,_threshold_otsu3,#ApplyThreshold#Any#,787

Before Change

            // using. Without a lower bound, you start having crazy threshold s that detect noise blobs. And same for
            // very crowded areas where there is zero background in the window. You want the foreground to be all
            // detected.
            t_min = max(self.threshold_range.min, t_global * 0.7)

            t_max = min(self.threshold_range.max, t_global * 1.5)

            t_local[t_local < t_min] = t_min

            t_local[t_local > t_max] = t_max

            return t_local, t_global

        t_local = t_global * self.threshold_correction_factor.value

        return min(max(t_local, self.threshold_range.min), self.threshold_range.max), t_global

    def _threshold_robust_background(self, image):
        average_fn = {
            RB_MEAN: numpy.mean,

After Change

        return skimage.img_as_float(t_local)

    def _threshold_otsu3(self, image):
        data = image.pixel_data

        mask = image.mask

        t_global = centrosome.threshold.get_otsu_threshold(
            data,
            mask,
            two_class_otsu=False,
            assign_middle_to_foreground=self.assign_middle_to_foreground.value == O_FOREGROUND
        )

        t_global = self._correct_global_threshold(t_global)

        if self.threshold_scope.value == TS_ADAPTIVE:
            if image.volumetric:
                t_local = numpy.zeros_like(data)

                for index, plane in enumerate(data):
                    t_local[index] = centrosome.threshold.get_adaptive_threshold(
                        centrosome.threshold.TM_OTSU,
                        plane,
                        t_global,
                        mask=mask[index],
                        adaptive_window_size=3,
                        two_class_otsu=False,
                        assign_middle_to_foreground=self.assign_middle_to_foreground.value == O_FOREGROUND
                    )
            else:
                t_local = centrosome.threshold.get_adaptive_threshold(
                    centrosome.threshold.TM_OTSU,
                    data,
                    t_global,
                    mask=mask,
                    adaptive_window_size=3,
                    two_class_otsu=False,
                    assign_middle_to_foreground=self.assign_middle_to_foreground.value == O_FOREGROUND
                )

            t_local = self._correct_local_threshold(t_local, t_global)

            return t_local, t_global

        return t_global, t_global

    def _threshold_robust_background(self, image):
        average_fn = {
            RB_MEAN: numpy.mean,

Instances