def make_image(self, magnification=1.0):
        if self._A is None:
            raise RuntimeError('You must first set the image array')

        x = self.to_rgba(self._A, self._alpha)
        self.magnification = magnification
        # if magnification is not one, we need to resize
        ismag = magnification!=1
        #if ismag: raise RuntimeError
        if ismag:
            isoutput = 0
        else:
            isoutput = 1
        im = _image.fromarray(x, isoutput)
        fc = self.figure.get_facecolor()
        im.set_bg( *mcolors.colorConverter.to_rgba(fc, 0) )
        im.is_grayscale = (self.cmap.name == "gray" and
                           len(self._A.shape) == 2)

        if ismag:
            numrows, numcols = self.get_size()
            numrows *= magnification
            numcols *= magnification
            im.set_interpolation(_image.NEAREST)
            im.resize(numcols, numrows)
        if self.origin=='upper':
            im.flipud_out()

        return im

    def stop_filter(self, post_processing):
        return RendererBase.stop_filter(self, post_processing)
        """
        Save the plot in the current canvas as a image and apply
        the *post_processing* function.

           def post_processing(image, dpi):
             # ny, nx, depth = image.shape
             # image (numpy array) has RGBA channels and has a depth of 4.
             ...
             # create a new_image (numpy array of 4 channels, size can be
             # different). The resulting image may have offsets from
             # lower-left corner of the original image
             return new_image, offset_x, offset_y

        The saved renderer is restored and the returned image from
        post_processing is plotted (using draw_image) on it.
        """

        # WARNING.
        # For agg_filter to work, the rendere's method need
        # to overridden in the class. See draw_markers, and draw_path_collections

        from matplotlib._image import fromarray

        width, height = int(self.width), int(self.height)

        buffer, bounds = self._renderer.tostring_rgba_minimized()

        l, b, w, h = bounds


        self._renderer = self._filter_renderers.pop()
        self._update_methods()

        if w > 0 and h > 0:
            img = np.fromstring(buffer, np.uint8)
            img, ox, oy = post_processing(img.reshape((h, w, 4)) / 255.,
                                          self.dpi)
            image = fromarray(np.flipud(img), 1)
#            image.flipud_out()

            gc = self.new_gc()
            self._renderer.draw_image(gc,
                                      l+ox, height - b - h +oy,
                                      image)

    def draw_tex(self, gc, x, y, s, prop, angle):
        # todo, handle props, angle, origins
        rgb = gc.get_rgb()
        size = prop.get_size_in_points()
        dpi = self.dpi.get()

        flip = angle==90
        w,h,d = self.get_text_width_height_descent(s, prop, 'TeX', rgb)
        if flip:
            w,h = h,w
            x -= w

        texmanager = self.get_texmanager()
        key = s, size, dpi, rgb, angle, texmanager.get_font_config()
        im = self.texd.get(key)
        if im is None:
            Z = texmanager.get_rgba(s, size, dpi, rgb)
            if flip:
                r = Z[:,:,0]
                g = Z[:,:,1]
                b = Z[:,:,2]
                a = Z[:,:,3]
                m,n,tmp = Z.shape

                def func(x):
                    return npy.transpose(npy.fliplr(x))

                Z = npy.zeros((n,m,4), float)
                Z[:,:,0] = func(r)
                Z[:,:,1] = func(g)
                Z[:,:,2] = func(b)
                Z[:,:,3] = func(a)
            im = fromarray(Z, 1)
            im.flipud_out()
            self.texd[key] = im

        cliprect = gc.get_clip_rectangle()
        if cliprect is None: bbox = None
        else: bbox = lbwh_to_bbox(*cliprect)
        self.draw_image(x, self.height-y, im, bbox)