Re: [Mypaint-discuss] Ordered dithering?
I've just implemented
A) a better test for roundtripping (slow, though.)
and
B) quick rounding tuning in pixops.hpp via macro use.
because of complicated git history I can't provide a patch, instead
I've attached the amended files.
According to this test, we currently badly roundtrip in approximately
half of the 256 possible cases. Most but not all of these errors are
truncation errors.
Changing the rounding behaviour might improve this. I haven't
managed to figure out how to get the desired behaviour yet, but the
new macros make the conversion formulae easy to tweak.
/* This file is part of MyPaint.
* Copyright (C) 2008-2009 by Martin Renold martin...@gmx.ch
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*/
// downscale a tile to half its size using bilinear interpolation
// used mainly for generating background mipmaps
void tile_downscale_rgb16(PyObject *src, PyObject *dst, int dst_x, int dst_y) {
#ifdef HEAVY_DEBUG
assert(PyArray_DIM(src, 0) == TILE_SIZE);
assert(PyArray_DIM(src, 1) == TILE_SIZE);
assert(PyArray_TYPE(src) == NPY_UINT16);
assert(PyArray_ISCARRAY(src));
assert(PyArray_TYPE(dst) == NPY_UINT16);
assert(PyArray_ISCARRAY(dst));
#endif
PyArrayObject* src_arr = ((PyArrayObject*)src);
PyArrayObject* dst_arr = ((PyArrayObject*)dst);
for (int y=0; yTILE_SIZE/2; y++) {
uint16_t * src_p = (uint16_t*)(src_arr-data + (2*y)*src_arr-strides[0]);
uint16_t * dst_p = (uint16_t*)(dst_arr-data + (y+dst_y)*dst_arr-strides[0]);
dst_p += 3*dst_x;
for(int x=0; xTILE_SIZE/2; x++) {
dst_p[0] = src_p[0]/4 + (src_p+3)[0]/4 + (src_p+3*TILE_SIZE)[0]/4 + (src_p+3*TILE_SIZE+3)[0]/4;
dst_p[1] = src_p[1]/4 + (src_p+3)[1]/4 + (src_p+3*TILE_SIZE)[1]/4 + (src_p+3*TILE_SIZE+3)[1]/4;
dst_p[2] = src_p[2]/4 + (src_p+3)[2]/4 + (src_p+3*TILE_SIZE)[2]/4 + (src_p+3*TILE_SIZE+3)[2]/4;
src_p += 6;
dst_p += 3;
}
}
}
// downscale a tile to half its size using bilinear interpolation
// used mainly for generating tiledsurface mipmaps
void tile_downscale_rgba16(PyObject *src, PyObject *dst, int dst_x, int dst_y) {
#ifdef HEAVY_DEBUG
assert(PyArray_DIM(src, 0) == TILE_SIZE);
assert(PyArray_DIM(src, 1) == TILE_SIZE);
assert(PyArray_TYPE(src) == NPY_UINT16);
assert(PyArray_ISCARRAY(src));
assert(PyArray_DIM(dst, 0) == TILE_SIZE);
assert(PyArray_DIM(dst, 1) == TILE_SIZE);
assert(PyArray_TYPE(dst) == NPY_UINT16);
assert(PyArray_ISCARRAY(dst));
#endif
PyArrayObject* src_arr = ((PyArrayObject*)src);
PyArrayObject* dst_arr = ((PyArrayObject*)dst);
for (int y=0; yTILE_SIZE/2; y++) {
uint16_t * src_p = (uint16_t*)(src_arr-data + (2*y)*src_arr-strides[0]);
uint16_t * dst_p = (uint16_t*)(dst_arr-data + (y+dst_y)*dst_arr-strides[0]);
dst_p += 4*dst_x;
for(int x=0; xTILE_SIZE/2; x++) {
dst_p[0] = src_p[0]/4 + (src_p+4)[0]/4 + (src_p+4*TILE_SIZE)[0]/4 + (src_p+4*TILE_SIZE+4)[0]/4;
dst_p[1] = src_p[1]/4 + (src_p+4)[1]/4 + (src_p+4*TILE_SIZE)[1]/4 + (src_p+4*TILE_SIZE+4)[1]/4;
dst_p[2] = src_p[2]/4 + (src_p+4)[2]/4 + (src_p+4*TILE_SIZE)[2]/4 + (src_p+4*TILE_SIZE+4)[2]/4;
dst_p[3] = src_p[3]/4 + (src_p+4)[3]/4 + (src_p+4*TILE_SIZE)[3]/4 + (src_p+4*TILE_SIZE+4)[3]/4;
src_p += 8;
dst_p += 4;
}
}
}
void tile_composite_rgba16_over_rgb16(PyObject * src, PyObject * dst, float alpha) {
#ifdef HEAVY_DEBUG
assert(PyArray_DIM(src, 0) == TILE_SIZE);
assert(PyArray_DIM(src, 1) == TILE_SIZE);
assert(PyArray_DIM(src, 2) == 4);
assert(PyArray_TYPE(src) == NPY_UINT16);
assert(PyArray_ISCARRAY(src));
assert(PyArray_DIM(dst, 0) == TILE_SIZE);
assert(PyArray_DIM(dst, 1) == TILE_SIZE);
assert(PyArray_DIM(dst, 2) == 3);
assert(PyArray_TYPE(dst) == NPY_UINT16);
assert(PyArray_ISBEHAVED(dst));
#endif
PyArrayObject* dst_arr = ((PyArrayObject*)dst);
#ifdef HEAVY_DEBUG
assert(dst_arr-strides[1] == 3*sizeof(uint16_t));
assert(dst_arr-strides[2] == sizeof(uint16_t));
#endif
uint32_t opac = alpha * (115) + 0.5;
opac = CLAMP(opac, 0, 115);
if (opac == 0) return;
uint16_t * src_p = (uint16_t*)((PyArrayObject*)src)-data;
char * p = dst_arr-data;
for (int y=0; yTILE_SIZE; y++) {
uint16_t * dst_p = (uint16_t*) (p);
for (int x=0; xTILE_SIZE; x++) {
// resultAlpha = 1.0 (thus it does not matter if resultColor is premultiplied alpha or not)
// resultColor = topColor + (1.0 - topAlpha) * bottomColor
const uint32_t one_minus_topAlpha = (115) - src_p[3]*opac/(115);
dst_p[0] = ((uint32_t)src_p[0]*opac + one_minus_topAlpha*dst_p[0]) / (115);
dst_p[1] = ((uint32_t)src_p[1]*opac + one_minus_topAlpha*dst_p[1]) / (115);
dst_p[2] = ((uint32_t)src_p[2]*opac + one_minus_topAlpha*dst_p[2]) / (115);
src_p += 4;
dst_p += 3;
}
p +=
[Mypaint-discuss] Fwd: Students exploring user interface design
-- Forwarded message -- From: Matthew Jadud mja...@allegheny.edu Date: 12 September 2010 23:04 Subject: Students exploring user interface design To: Matthew Jadud mja...@allegheny.edu Hello all, I will keep my note brief. I'm writing you because you, or a colleague of yours, responded to my call regarding usability/interaction design[1]. I've tried to capture all of your projects in a list[2], and have done my best to (briefly) write something about your projects. I want to say that I am overwhelmed by the number of responses I received, and from so many projects that I am familiar with in one form or another. I'm very excited about the opportunities your projects represent for the students. The students are currently engaging in their first UI design and testing project, which involves student involvement with the Fedora Design community surrounding their pending website redesign. The students don't *really* know what they're doing yet, and Máirín Duffy and the websites team were kind enough to let the students use this as an opportunity to skin their knees a bit and learn what the whole UI design/test process looks like. As I've heard some FOSS practitioners say, the students are learning to be productively lost. While they're doing that, they're reading Norman, Cooper, Snyder, and other resources, delving more deeply into the cognitive and practical aspects of UI design. They're going to be looking at all of your projects over the coming week, and will likely begin contacting some of you next week to ask questions and express interest in working with you. In short, this note is a heads up as to what we've been up to, and where we're going next. I'm sorry I didn't write sooner, but I've been doing a lot of community interaction/bridging myself these past two weeks. The students are engaged, and the knowledge that they will be working with real projects has, I think, really focused their efforts so far. Please feel free to ask me questions. Mostly, I think you can continue to sit back and wait for students to come your way. So, again: thank you! Cheers, Matt PS. Given that we only have 10 students in the class, the fact is that we won't be able to work with most of the projects that responded. That said, I'm going to push to see if I can get this course cross-listed with our Psychology department, and perhaps get it to run more often. Or, perhaps I'll look at ways I can do something that blends our local instruction with P2PU. Either way, it is clear that there are a lot of great opportunities to introduce students to open communities around the theme of usability. [1] http://www.sububi.org/2010/08/18/free-interaction-design-for-your-open-source-project/ [2] http://rockalypse.org/courses/cmpsc303f10/projects/ [3] http://wiki.rockalypse.org/HCD -- Regards Jon Nordby - www.jonnor.com ___ Mypaint-discuss mailing list Mypaint-discuss@gna.org https://mail.gna.org/listinfo/mypaint-discuss
