[FFmpeg-devel] rectification filter
Daniel Oberhoff
danieloberhoff at gmail.com
Tue Aug 12 00:52:45 CEST 2014
Am 11.08.2014 um 17:39 schrieb Clément Bœsch <u at pkh.me>:
> On Mon, Aug 11, 2014 at 05:33:10PM +0200, Daniel Oberhoff wrote:
>>
>> Am 11.08.2014 um 00:20 schrieb Michael Niedermayer <michaelni at gmx.at>:
>>
>>> On Sun, Aug 10, 2014 at 08:36:37PM +0200, Daniel Oberhoff wrote:
>>>>
>>>>
>>>> Von meinem iPhone gesendet
>>>>
>>>>> Am 10.08.2014 um 17:50 schrieb Michael Niedermayer <michaelni at gmx.at>:
>>>>>
>>>>>> On Sun, Aug 10, 2014 at 03:36:04PM +0200, Nicolas George wrote:
>>>>>> Le tridi 23 thermidor, an CCXXII, Paul B Mahol a écrit :
>>>>>>> It is not mandatory(but it would be nice) to add other methods to have
>>>>>>> this filter included into libavfilter.
>>>>>>
>>>>>> Is it really a good idea? We would end up with various interpolation /
>>>>>> anti-aliasing algorithms implemented in each filter that needs it, none of
>>>>>> them with the same set of architecture optimizations and each with its own
>>>>>> set of bugs and misfeatures.
>>>>>
>>>>>> Until someone comes up with a really satisfactory solution, I believe the
>>>>>> simple solution of suggesting users to upscale before the filter and
>>>>>> downscale after, using the optimized lswr scalers, is better.
>>>>>
>>>>> thats not practical
>>>>> a 1024x768 image would need to be upscaled to 262144x196608 to get
>>>>> 8bit precission from a nearest neighbor resampler as basis
>>>>>
>>>>> also see vf_perspective.c which supports bilinear and bicubic
>>>>> interpolation, these surely could be shared and exist already.
>>>>>
>>>>> of course we could push it with just nearest neighbor and work on that
>>>>> later, but i dont think only nearest neighbor and leaving it at that
>>>>> is a reasonable choice, its too poor quality wise
>>>>
>>>> I would like to push now, and try to generate a refactoring patch possibly within the week.
>>>
>>> sure, feel free to push and send a pull request, or if you prefer
>>> i can allso apply the patch if you post the latest verision
>>
>> I am not sure about pull requests (would have to read up on how to do that properly) so rigjht now I would prefer it by patch:
>>
>>
>>
>> Is that ok?
>>
>
> Yes, please just re-submit the patch here.
Darn, forgot that attachements are scrubbed from the list (I had the patch attached). Will paste at the end.
>
> Sorry about the suggestion of adding FATE tests, I forgot the fact that
> the filter was actually using floats.
No probs :)
From aa27195163da74c6e8f6a3b258f971f589d19aca Mon Sep 17 00:00:00 2001
From: Daniel Oberhoff <daniel at danieloberhoff.de>
Date: Mon, 28 Jul 2014 23:58:12 +0200
Subject: [PATCH] avfilter: ported lenscorrection filter from frei0r
---
Changelog | 2 +-
MAINTAINERS | 1 +
doc/filters.texi | 46 ++++++++++
libavfilter/Makefile | 1 +
libavfilter/allfilters.c | 1 +
libavfilter/vf_lenscorrection.c | 199 ++++++++++++++++++++++++++++++++++++++++
6 files changed, 249 insertions(+), 1 deletion(-)
create mode 100644 libavfilter/vf_lenscorrection.c
diff --git a/Changelog b/Changelog
index 25dd210..79393ce 100644
--- a/Changelog
+++ b/Changelog
@@ -3,7 +3,7 @@ releases are sorted from youngest to oldest.
version <next>:
- Icecast protocol
-
+- ported lenscorrection filter from frei0r filter
version 2.3:
- AC3 fixed-point decoding
diff --git a/MAINTAINERS b/MAINTAINERS
index c383d3f..6948e22 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -354,6 +354,7 @@ Filters:
vf_histogram.c Paul B Mahol
vf_hqx.c Clément Bœsch
vf_il.c Paul B Mahol
+ vf_lenscorrection.c Daniel Oberhoff
vf_mergeplanes.c Paul B Mahol
vf_psnr.c Paul B Mahol
vf_scale.c Michael Niedermayer
diff --git a/doc/filters.texi b/doc/filters.texi
index e0759fc..54b4451 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -5546,6 +5546,51 @@ kerndeint=map=1
@end example
@end itemize
+ at section lenscorrection
+
+Correct radial lens distortion
+
+This filter can be used to correct for radial distortion as can result from the use
+of wide angle lenses, and thereby re-rectify the image. To find the right parameters
+one can use tools available for example as part of opencv or simply trial-and-error.
+To use opencv use the calibration sample (under samples/cpp) from the opencv sources
+and extract the k1 and k2 coefficients from the resulting matrix.
+
+Note that effectively the same filter is available in the open-source tools Krita and
+Digikam from the KDE project.
+
+In contrast to the @ref{vignette} filter, which can also be used to compensate lens errors,
+this filter corrects the distortion of the image, whereas @ref{vignette} corrects the
+brightness distribution, so you may want to use both filters together in certain
+cases, though you will have to take care of ordering, i.e. whether vignetting should
+be applied before or after lens correction.
+
+ at subsection Options
+
+The filter accepts the following options:
+
+ at table @option
+ at item cx
+Relative x-coordinate of the focal point of the image, and thereby the center of the
+distrortion. This value has a range [0,1] and is expressed as fractions of the image
+width.
+ at item cy
+Relative y-coordinate of the focal point of the image, and thereby the center of the
+distrortion. This value has a range [0,1] and is expressed as fractions of the image
+height.
+ at item k1
+Coefficient of the quadratic correction term. 0.5 means no correction.
+ at item k2
+Coefficient of the double quadratic correction term. 0.5 means no correction.
+ at end table
+
+The formula that generates the correction is:
+
+ at var{r_src} = @var{r_tgt} * (1 + @var{k1} * (@var{r_tgt} / @var{r_0})^2 + @var{k2} * (@var{r_tgt} / @var{r_0})^4)
+
+where @var{r_0} is halve of the image diagonal and @var{r_src} and @var{r_tgt} are the
+distances from the focal point in the source and target images, respectively.
+
@anchor{lut3d}
@section lut3d
@@ -8758,6 +8803,7 @@ For example, to vertically flip a video with @command{ffmpeg}:
ffmpeg -i in.avi -vf "vflip" out.avi
@end example
+ at anchor{vignette}
@section vignette
Make or reverse a natural vignetting effect.
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 0f54381..e9c8456 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -138,6 +138,7 @@ OBJS-$(CONFIG_IL_FILTER) += vf_il.o
OBJS-$(CONFIG_INTERLACE_FILTER) += vf_interlace.o
OBJS-$(CONFIG_INTERLEAVE_FILTER) += f_interleave.o
OBJS-$(CONFIG_KERNDEINT_FILTER) += vf_kerndeint.o
+OBJS-$(CONFIG_LENSCORRECTION_FILTER) += vf_lenscorrection.o
OBJS-$(CONFIG_LUT3D_FILTER) += vf_lut3d.o
OBJS-$(CONFIG_LUT_FILTER) += vf_lut.o
OBJS-$(CONFIG_LUTRGB_FILTER) += vf_lut.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 1877557..b1d6ff5 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -156,6 +156,7 @@ void avfilter_register_all(void)
REGISTER_FILTER(INTERLACE, interlace, vf);
REGISTER_FILTER(INTERLEAVE, interleave, vf);
REGISTER_FILTER(KERNDEINT, kerndeint, vf);
+ REGISTER_FILTER(LENSCORRECTION, lenscorrection, vf);
REGISTER_FILTER(LUT3D, lut3d, vf);
REGISTER_FILTER(LUT, lut, vf);
REGISTER_FILTER(LUTRGB, lutrgb, vf);
diff --git a/libavfilter/vf_lenscorrection.c b/libavfilter/vf_lenscorrection.c
new file mode 100644
index 0000000..c2346db
--- /dev/null
+++ b/libavfilter/vf_lenscorrection.c
@@ -0,0 +1,199 @@
+/*
+ * Copyright (C) 2007 Richard Spindler (author of frei0r plugin from which this was derived)
+ * Copyright (C) 2014 Daniel Oberhoff
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Lenscorrection filter, algorithm from the frei0r plugin with the same name
+*/
+#include <stdlib.h>
+#include <math.h>
+
+#include "libavutil/opt.h"
+#include "libavutil/intreadwrite.h"
+#include "libavutil/pixdesc.h"
+
+#include "avfilter.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct LenscorrectionCtx {
+ const AVClass *av_class;
+ unsigned int width;
+ unsigned int height;
+ int hsub, vsub;
+ int nb_planes;
+ double cx, cy, k1, k2;
+} LenscorrectionCtx;
+
+#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
+static const AVOption lenscorrection_options[] = {
+ { "cx", "set relative center x", offsetof(LenscorrectionCtx, cx), AV_OPT_TYPE_DOUBLE, {.dbl=0.5}, 0, 1, .flags=FLAGS },
+ { "cy", "set relative center y", offsetof(LenscorrectionCtx, cy), AV_OPT_TYPE_DOUBLE, {.dbl=0.5}, 0, 1, .flags=FLAGS },
+ { "k1", "set quadratic distortion factor", offsetof(LenscorrectionCtx, k1), AV_OPT_TYPE_DOUBLE, {.dbl=0.0}, -1, 1, .flags=FLAGS },
+ { "k2", "set double quadratic distortion factor", offsetof(LenscorrectionCtx, k2), AV_OPT_TYPE_DOUBLE, {.dbl=0.0}, -1, 1, .flags=FLAGS },
+ { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(lenscorrection);
+
+typedef struct ThreadData {
+ AVFrame *in, *out;
+ float w, h;
+ int plane;
+ float xcenter, ycenter;
+ float k1, k2;
+} ThreadData;
+
+static int filter_slice(AVFilterContext *ctx, void *arg, int job, int nb_jobs)
+{
+ ThreadData *td = (ThreadData*)arg;
+ AVFrame *in = td->in;
+ AVFrame *out = td->out;
+
+ const float w = td->w, h = td->h;
+ const float xcenter = td->xcenter;
+ const float ycenter = td->ycenter;
+ const float r2inv = 4.0 / (w * w + h * h);
+ const float k1 = td->k1;
+ const float k2 = td->k2;
+ const int start = (h * job ) / nb_jobs;
+ const int end = (h * (job+1)) / nb_jobs;
+ const int plane = td->plane;
+ const int inlinesize = in->linesize[plane];
+ const int outlinesize = out->linesize[plane];
+ const uint8_t *indata = in->data[plane];
+ uint8_t *outrow = out->data[plane] + start * outlinesize;
+ int i;
+ for (i = start; i < end; i++, outrow += outlinesize) {
+ const float off_y = i - ycenter;
+ const float off_y2 = off_y * off_y;
+ uint8_t *out = outrow;
+ int j;
+ for (j = 0; j < w; j++) {
+ const float off_x = j - xcenter;
+ const float r2 = (off_x * off_x + off_y2) * r2inv;
+ const float radius_mult = 1.0f + r2 * k1 + r2 * r2 * k2;
+ const int x = xcenter + radius_mult * off_x + 0.5f;
+ const int y = ycenter + radius_mult * off_y + 0.5f;
+ const char isvalid = x > 0 && x < w - 1 && y > 0 && y < h - 1;
+ *out++ = isvalid ? indata[y * inlinesize + x] : 0;
+ }
+ }
+ return 0;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+ static enum AVPixelFormat pix_fmts[] = {
+ AV_PIX_FMT_YUV410P,
+ AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P,
+ AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVJ420P,
+ AV_PIX_FMT_YUVA444P, AV_PIX_FMT_YUVA420P,
+ AV_PIX_FMT_YUV422P,
+ AV_PIX_FMT_NONE
+ };
+
+ ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
+ return 0;
+}
+
+static int config_props(AVFilterLink *outlink)
+{
+ AVFilterContext *ctx = outlink->src;
+ LenscorrectionCtx *rect = ctx->priv;
+ AVFilterLink *inlink = ctx->inputs[0];
+ const AVPixFmtDescriptor *pixdesc = av_pix_fmt_desc_get(inlink->format);
+ rect->hsub = pixdesc->log2_chroma_w;
+ rect->vsub = pixdesc->log2_chroma_h;
+ outlink->w = rect->width = inlink->w;
+ outlink->h = rect->height = inlink->h;
+ rect->nb_planes = av_pix_fmt_count_planes(inlink->format);
+ return 0;
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+ AVFilterContext *ctx = inlink->dst;
+ AVFilterLink *outlink = ctx->outputs[0];
+ LenscorrectionCtx *rect = (LenscorrectionCtx*)ctx->priv;
+ AVFrame *out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+ int plane;
+
+ if (!out) {
+ av_frame_free(&in);
+ return AVERROR(ENOMEM);
+ }
+
+ av_frame_copy_props(out, in);
+
+ for (plane = 0; plane < rect->nb_planes; ++plane) {
+ int hsub = plane == 1 || plane == 2 ? rect->hsub : 0;
+ int vsub = plane == 1 || plane == 2 ? rect->vsub : 0;
+ float hdiv = 1 << hsub;
+ float vdiv = 1 << vsub;
+ float w = rect->width / hdiv;
+ float h = rect->height / vdiv;
+ ThreadData td = {
+ .in = in,
+ .out = out,
+ .w = w,
+ .h = h,
+ .xcenter = rect->cx * w,
+ .ycenter = rect->cy * h,
+ .k1 = rect->k1,
+ .k2 = rect->k2,
+ .plane = plane};
+ ctx->internal->execute(ctx, filter_slice, &td, NULL, FFMIN(h, ctx->graph->nb_threads));
+ }
+
+ av_frame_free(&in);
+ return ff_filter_frame(outlink, out);
+}
+
+static const AVFilterPad lenscorrection_inputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .filter_frame = filter_frame,
+ },
+ { NULL }
+};
+
+static const AVFilterPad lenscorrection_outputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .config_props = config_props,
+ },
+ { NULL }
+};
+
+AVFilter ff_vf_lenscorrection = {
+ .name = "lenscorrection",
+ .description = NULL_IF_CONFIG_SMALL("Rectify the image by correcting for lens distortion."),
+ .priv_size = sizeof(LenscorrectionCtx),
+ .query_formats = query_formats,
+ .inputs = lenscorrection_inputs,
+ .outputs = lenscorrection_outputs,
+ .priv_class = &lenscorrection_class,
+ .flags = AVFILTER_FLAG_SLICE_THREADS,
+};
+
\ No newline at end of file
--
1.8.2
More information about the ffmpeg-devel
mailing list