[NTG-context] Downsampling images (again)

[NTG-context] Downsampling images (again)

[Henning Hraban Ramm]

This is lying around for years, let’s try to get it finally working.
The code below works, but doesn’t do the right thing.

Since the original image doesn’t contain a resolution setting, gm 
doesn’t do anything:
“gm convert: image does not contain resolution”

It would be more reliable to use “gm convert -resize”, but for that we 
need to know the final pixel size.

How can a converter function access …
– the final (scaled) size of a placed image
– the original pixel size of an image
?
With that information it would be easy to calculate the target pixel size.

And then it would be nice to hook this function into 
\setupexternalfigure[conversion=]


"""
\startluacode
local function downsampler(oldname, newname, resolution)
     if not resolution or resolution == "" then
         resolution = 72
     end
     local cmd = string.format(
       [[gm convert -resample %ix%i "%s" "%s"]],
       resolution, resolution, oldname, newname)
     logs.report("downsample", cmd)
     os.execute(cmd)
end

-- Set the PDF and default JPEG converters to the above function.
figures.converters.jpg.pdf = downsampler
figures.converters.jpg.default = downsampler
\stopluacode

\setupexternalfigure[
   %conversion=downsampler,% not used
   resolution=10,
]

\starttext

\externalfigure[https://upload.wikimedia.org/wikipedia/commons/d/dd/Hermann_Zapf_signing.jpg][width=.5\textwidth]

\stoptext
"""

Hraban
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Alan Braslau via ntg-context]

Hraban,

I would first "fix" my image sources, individually, to contain missing
resolution information.

Alan

On Sat, 21 Sep 2024 13:21:08 +0200
Henning Hraban Ramm <texml@fiee.net> wrote:

> This is lying around for years, let’s try to get it finally working.
> The code below works, but doesn’t do the right thing.
> 
> Since the original image doesn’t contain a resolution setting, gm 
> doesn’t do anything:
> “gm convert: image does not contain resolution”
> 
> It would be more reliable to use “gm convert -resize”, but for that
> we need to know the final pixel size.
> 
> How can a converter function access …
> – the final (scaled) size of a placed image
> – the original pixel size of an image
> ?
> With that information it would be easy to calculate the target pixel
> size.
> 
> And then it would be nice to hook this function into 
> \setupexternalfigure[conversion=]
> 
> 
> """
> \startluacode
> local function downsampler(oldname, newname, resolution)
>      if not resolution or resolution == "" then
>          resolution = 72
>      end
>      local cmd = string.format(
>        [[gm convert -resample %ix%i "%s" "%s"]],
>        resolution, resolution, oldname, newname)
>      logs.report("downsample", cmd)
>      os.execute(cmd)
> end
> 
> -- Set the PDF and default JPEG converters to the above function.
> figures.converters.jpg.pdf = downsampler
> figures.converters.jpg.default = downsampler
> \stopluacode
> 
> \setupexternalfigure[
>    %conversion=downsampler,% not used
>    resolution=10,
> ]
> 
> \starttext
> 
> \externalfigure[https://upload.wikimedia.org/wikipedia/commons/d/dd/Hermann_Zapf_signing.jpg][width=.5\textwidth]
> 
> \stoptext
> """
> 
> Hraban
> ___________________________________________________________________________________
> If your question is of interest to others as well, please add an
> entry to the Wiki!
> 
> maillist : ntg-context@ntg.nl
> / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
> archive  : https://github.com/contextgarden/context
> wiki     : https://wiki.contextgarden.net
> ___________________________________________________________________________________
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Henning Hraban Ramm]

Am 21.09.24 um 16:30 schrieb Alan Braslau:
> I would first "fix" my image sources, individually, to contain missing
> resolution information.

The resolution setting is just metadata, it doesn’t affect pixel or file 
size. I need a function that works with “real” (pixel) sizes of images.

Hraban
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Alan Braslau via ntg-context]

On Sat, 21 Sep 2024 21:44:53 +0200
Henning Hraban Ramm <texml@fiee.net> wrote:

> Am 21.09.24 um 16:30 schrieb Alan Braslau:
> > I would first "fix" my image sources, individually, to contain
> > missing resolution information.  
> 
> The resolution setting is just metadata, it doesn’t affect pixel or
> file size. I need a function that works with “real” (pixel) sizes of
> images.

Exactly. The raw data is just pixels, and metadata sets what the
expected pixel size might be. Of course, this metadata can be
meaningless as the user can and often does rescale the image to meet
the typesetting needs.

The image manipulation tools use this metadata to resample resolution.
I found this very useful as I used *hundreds* of scanned X-ray images
that had real sizes, high resolution and 16 or even 24 bit depth. Of
course, I had to reduce all of this in order to include them in a
digestible and printable pdf.

Alan
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Peter Münster]

[-- Attachment #1: Type: text/plain, Size: 310 bytes --]

On Sat, Sep 21 2024, Henning Hraban Ramm wrote:

> How can a converter function access …
> – the final (scaled) size of a placed image

Hi,

I use "figures.current().request" in my module (here attached).


> – the original pixel size of an image ?

I use img.scan() for that.

HTH,
-- 
           Peter

[-- Attachment #2: grph-downsample.lua --]
[-- Type: application/octet-stream, Size: 1610 bytes --]

if not modules then modules = { } end modules ['grph-downsample'] = {
  version   = 1.001,
  comment   = "companion to grph-inc.mkiv",
  author    = "Peter Münster",
  copyright = "PRAGMA ADE / ConTeXt Development Team",
  license   = "see context related readme files"
}

local format = string.format
-- figures.cachepaths.path = "cache" -- should be setup-option
local function sample_down(oldname, newname, resolution)
    local request = figures.current().request
    local width = request.width
    local height = request.height
    if resolution == "" or (not width and not height) then
        print(format("Nothing to do: %s, %s, %s", oldname, newname, resolution))
        return
    end
    local inch = 72.27
    local image = img.scan{filename = oldname}
    local xy = image.xsize / image.ysize
    if not width then
        width = height * xy / 65536
    end
    if not height then
        height = width / xy / 65536
    end
    local xsize = resolution * width / inch
    local ysize = resolution * height / inch
    if xsize < image.xsize or ysize < image.ysize then
        local s = format("gm convert -resize %dx%d %s %s",
                         xsize, ysize, oldname, newname)
        print("Conversion: " .. s)
        os.execute(s)
    else
        print(format("Nothing to do: %s, %s, %s", oldname, newname, resolution))
        print(format("xsize = %d, ysize = %d", xsize, ysize))
    end
end

local formats = {"png", "jpg", "gif"}

for _, s in ipairs(formats) do
    figures.converters[s] = figures.converters[s] or {}
    figures.converters[s]["lowres." .. s] = sample_down
end

[-- Attachment #3: Type: text/plain, Size: 511 bytes --]

___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Henning Hraban Ramm]

Am 22.09.24 um 09:06 schrieb Peter Münster:
> On Sat, Sep 21 2024, Henning Hraban Ramm wrote:
> 
>> How can a converter function access …
>> – the final (scaled) size of a placed image
> 
> Hi,
> 
> I use "figures.current().request" in my module (here attached).
> 
> 
>> – the original pixel size of an image ?
> 
> I use img.scan() for that.

Hi Peter,
sorry, I forgot to answer.
I was using your code for years, and AFAIR it also doesn’t use the final 
scaled size. But I must check again – too much going on ATM…
Hraban
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Henning Hraban Ramm]

[-- Attachment #1: Type: text/plain, Size: 1192 bytes --]

Am 22.09.24 um 09:06 schrieb Peter Münster:
> On Sat, Sep 21 2024, Henning Hraban Ramm wrote:
> 
>> How can a converter function access …
>> – the final (scaled) size of a placed image
> 
> Hi,
> 
> I use "figures.current().request" in my module (here attached).
> 
> 
>> – the original pixel size of an image ?
> 
> I use img.scan() for that.

Hi Peter,
are you sure that your code works with LMTX?

Find attached my version.
It seems to work, but only for the first instance of an image, not for 
different sizes of the same image – makes sense WRT efficiency, but 
might give wrong results for some uses.

Here’s my test code:

"""
\loadluafile[grph-downsample.lua]

\setupexternalfigure[
   location={local,global,default},
   conversion=lowres.jpg,
   resolution=20,
]

\starttext

\useexternalfigure[zapf][https://upload.wikimedia.org/wikipedia/commons/d/dd/Hermann_Zapf_signing.jpg]
\useexternalfigure[lino][https://upload.wikimedia.org/wikipedia/commons/b/bd/Linotype-vorne-deutsches-museum.jpg]

\externalfigure[zapf][width=.5tw]

\externalfigure[lino][height=.3th]

\externalfigure[zapf][height=.5th]

\stoptext
"""

Hraban

[-- Attachment #2: grph-downsample.lua --]
[-- Type: text/plain, Size: 2811 bytes --]

if not modules then modules = { } end modules ['grph-downsample'] = {
  version   = 1.103,
  comment   = "companion to grph-inc.mkiv",
  author    = "Peter Münster", -- adapted to LMTX by Hraban
  copyright = "PRAGMA ADE / ConTeXt Development Team",
  license   = "see context related readme files"
}

local format = string.format
local report = logs.reporter("DOWNSAMPLE")
local function sample_down(oldname, newname, resolution)
  report("sample_down %s to %s (%s)", oldname, newname, resolution)
  local request = figures.current().request
  -- this is probably the original pixel size
  local width = request.width
  local height = request.height
  -- width/height are integer in sp or a boolean
  local TEXpt = 65536
  local inch = 72.27
  if resolution == "" then -- or (not width and not height) then
    report("Nothing to do (missing resolution): %s, %s, %s dpi, %s x %s px", oldname, newname, resolution, width, height)
    return
  else
    report("Requested: %s, %s, %s dpi, %s x %s px", oldname, newname, resolution, width, height)
  end

  -- MkIV:
  -- local image = img.scan{filename = oldname}

  -- LMTX:
  local image = figures.getinfo(oldname,1)
  image = image.status.private

  local xy = image.xsize / image.ysize -- size ratio
  if (not width and not height) then
    -- no size requested? use default width
    -- probably wrong…
    width = 300 * TEXpt
  end
  if not width then
    height = height / TEXpt
    width = height * xy
  end
  if not height then
    width = width / TEXpt
    height = width / xy
  end
  width = math.floor(width)
  height = math.floor(height)
  report("image size %dx%dpx. requested %dx%dpt at %ddpi.", image.xsize, image.ysize, width, height, resolution)
  local xsize = math.floor(resolution * width / inch)
  local ysize = math.floor(resolution * height / inch)
  report("resize %dx%dpx to %dx%dpx:", image.xsize, image.ysize, xsize, ysize)
  if xsize < image.xsize or ysize < image.ysize then
    -- maybe add a tolerance of 20% (don’t downsample if the difference is marginal)
    local s = format("gm convert -resize %dx%d -resample %dx%d \"%s\" \"%s\"",
                     xsize, ysize, resolution, resolution, oldname, newname)
    report("calling: %s", s)
    os.execute(s)
  else
    report("Nothing to do (image smaller than requested): %s, %s, %s dpi, %d x %d px", oldname, newname, resolution, width, height)
    report("xsize = %d, ysize = %d", xsize, ysize)
  end
end

local formats = {"png", "jpg", "gif"}

for _, s in ipairs(formats) do
  figures.converters[s] = figures.converters[s] or {}
  -- use \setupexternalfigure[conversion=lowres.jpg]:
  figures.converters[s]["lowres." .. s] = sample_down
  -- use without setup:
  --figures.converters[s].default = sample_down
  --figures.converters[s].pdf = sample_down -- gets used
end

[-- Attachment #3: Type: text/plain, Size: 511 bytes --]

___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________

[NTG-context] Re: Downsampling images (again)

[Peter Münster]

[-- Attachment #1: Type: text/plain, Size: 549 bytes --]

On Wed, Oct 16 2024, Henning Hraban Ramm wrote:

>> I use img.scan() for that.
>
> are you sure that your code works with LMTX?

Hi Hraban,

No, sorry, I didn’t test it...


> Find attached my version.

Thanks, I have integrated your ideas in a new version here attached.


> It seems to work, but only for the first instance of an image,

Yes, you need to patch the file grph-inc.lmt:
https://lists.contextgarden.net/archives/list/ntg-context@ntg.nl/message/ES7Q57MU5UIEV2B2AWEAPR66MMXI6ZHY/

Cheers,
-- 
           Peter

[-- Attachment #2: grph-downsample.lua --]
[-- Type: application/octet-stream, Size: 2299 bytes --]

if not modules then modules = { } end modules ['grph-downsample'] = {
  version   = 1.104,
  comment   = "companion to grph-inc.mkiv",
  author    = "Peter Münster and Henning Hraban Ramm",
  copyright = "PRAGMA ADE / ConTeXt Development Team",
  license   = "see context related readme files"
}

-- todo:
-- figures.cachepaths.path = "cache" -- should be a setup-option
-- threshold should be a setup-option

local function round(x)
    return math.floor(x + 0.5)
end

local format = string.format
local report = logs.reporter("DOWNSAMPLE")

local function sample_down(oldname, newname, resolution)
    report("sample_down %s to %s (%s)", oldname, newname, resolution)
    local request = figures.current().request
    local width   = request.width
    local height  = request.height
    if resolution == "" or (not width and not height) then
        report("Nothing to do: %s, %s, %s", oldname, newname, resolution)
        return
    else
        report("Requested: %s, %s, %s dpi, %s sp x %s sp", oldname, newname,
               resolution, width or "???", height or "???")
    end
    -- requested size must be smaller than original size
    -- multiplied by this threshold:
    local threshold = 0.8
    local inch      = 72.27 * 65536   -- an inch in scaled points
    local image     = figures.getinfo(oldname, 1)
    image = image.status.private

    local xy = image.xsize / image.ysize
    if not width then
        width = height * xy
    end
    if not height then
        height = width / xy
    end
    local xsize = round(resolution * width  / inch)
    local ysize = round(resolution * height / inch)
    report("resize %d x %d to %d x %d:", image.xsize, image.ysize,
           xsize, ysize)
    if xsize < image.xsize * threshold or ysize < image.ysize * threshold then
        local s = format("gm convert -resize %dx%d %s %s",
                         xsize, ysize, oldname, newname)
        report("calling: %s", s)
        os.execute(s)
    else
        report("Nothing to do: %s, %s, %s", oldname, newname, resolution)
        report("xsize = %d, ysize = %d", xsize, ysize)
    end
end

local formats = {"png", "jpg", "gif"}

for _, s in ipairs(formats) do
    figures.converters[s] = figures.converters[s] or {}
    figures.converters[s]["lowres." .. s] = sample_down
end

[-- Attachment #3: Type: text/plain, Size: 511 bytes --]

___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the Wiki!

maillist : ntg-context@ntg.nl / https://mailman.ntg.nl/mailman3/lists/ntg-context.ntg.nl
webpage  : https://www.pragma-ade.nl / https://context.aanhet.net (mirror)
archive  : https://github.com/contextgarden/context
wiki     : https://wiki.contextgarden.net
___________________________________________________________________________________