How does pillow implement Transform.QUAD and Transform.MESH

[kyrollosyanny]

Hello,

I am trying to understand how Pillow implements the Transform.MESH method used in ImageOps.deform. In the source code I found the following

        if method == MESH:
            # list of quads
            for box, quad in data:
                im.__transformer(box, self, QUAD, quad, resample, fill)
        else:
            im.__transformer((0, 0)+size, self, method, data, resample, fill)

So for each entry in the list of quads, it calls the __transformer function with method=QUAD. Then in the __transformer function it performs some math on the quad and box data and calls self.im.transform2(box, image.im, method, data, resample, fill) which I don't seem to find a function for or understand how it does the QUAD transform.

Any documentation or pointers to source good will be very helpful.

Thank you

[Yay295]

transform2 is written in C, and was renamed to transform a few months ago.

Pillow/src/_imaging.c

Line 3630 in 7097a9a

{"transform", (PyCFunction)_transform, METH_VARARGS},

Pillow/src/_imaging.c

Lines 2010 to 2072 in 7097a9a

	static PyObject *
	_transform(ImagingObject *self, PyObject *args) {
	static const char *wrong_number = "wrong number of matrix entries";
	Imaging imOut;
	Py_ssize_t n;
	double *a;
	ImagingObject *imagep;
	int x0, y0, x1, y1;
	int method;
	PyObject *data;
	int filter = IMAGING_TRANSFORM_NEAREST;
	int fill = 1;
	if (!PyArg_ParseTuple(
	args,
	"(iiii)O!iO|ii",
	&x0,
	&y0,
	&x1,
	&y1,
	&Imaging_Type,
	&imagep,
	&method,
	&data,
	&filter,
	&fill
	)) {
	return NULL;
	}
	switch (method) {
	case IMAGING_TRANSFORM_AFFINE:
	n = 6;
	break;
	case IMAGING_TRANSFORM_PERSPECTIVE:
	n = 8;
	break;
	case IMAGING_TRANSFORM_QUAD:
	n = 8;
	break;
	default:
	n = -1; /* force error */
	}
	a = getlist(data, &n, wrong_number, TYPE_DOUBLE);
	if (!a) {
	return NULL;
	}
	imOut = ImagingTransform(
	self->image, imagep->image, method, x0, y0, x1, y1, a, filter, fill
	);
	free(a);
	if (!imOut) {
	return NULL;
	}
	Py_INCREF(Py_None);
	return Py_None;
	}

[kyrollosyanny]

Got it! Thank you so much