pamperspective

Updated: 2 September 2004
Table Of Contents

NAME

pamperspective - a reverse scanline renderer for Netpbm images

SYNOPSIS

pamperspective 
    [--bottom_margin=num]
    [--detail=num]
    [--frame_include=bool]
    [--height=num]
    [--include=[x1,y1;x2,y2; ...]]
    [--input_system=spec]
    [--input_unit=spec]
    [--interpolation=spec]
    [--left_margin=num]
    [--margin=num]
    [--output_system=spec]
    [--proportion=spec]
    [--ratio=num]
    [--right_margin=num]
    [--top_margin=num]
    [--width=num]
    {
      {
        upper_left_x upper_left_y upper_right_x upper_right_y
        lower_left_x lower_left_y lower_right_x lower_right_y
      }
      |
      {
        {--upper_left_x|--ulx}=upper_left_x
        {--upper_left_y|--uly}=upper_left_y
        {--upper_right_x|--urx}=upper_right_x
        {--upper_right_y|--ury}=upper_right_y
        {--lower_left_x|--llx}=lower_left_x
        {--lower_left_y|--lly}=lower_left_y
        {--lower_right_x|--lrx}=lower_right_x
        {--lower_right_y|--lry}=lower_right_y
      }
   }
   [infile]

Minimum unique abbreviation of option is acceptable. (But note that shortest unique prefixes might be longer in future versions of the program.) You may use single hyphens instead of double hyphen to denote options. You may use white space in place of the equals sign to separate an option name from its value. All options starting with hyphens may be given in any order.

DESCRIPTION

This program is part of Netpbm.

pamperspective reads a Netpbm image as input and produces a Netpbm image of the same format as output.

pamperspective interprets the input image as a perspective projection of another image which is in a plane oblique to that of the input image. For example, a photograph of a painting, taken at an angle. The arguments upper_left_x ... lower_right_y specify a quadrilateral in the photograph that pamperspective assumes corresponds to a parallelogram in the painting. The output image consists of this parallelogram, sheared to a rectangle. In this way pamperspective undoes the effect of a raytracer or scanline renderer.

Note that if the input image is a projection of a solid scene, rather than a plane, the result is like a different camera angle on that scene, to the extent that the scene is shallow from the other angle.

The input is from infile, or from Standard Input, if infile is not specified. The output is to Standard Output.

OPTIONS

For options of the form --name=num, You can specify the value num in any of the traditional ways. Additionally, you can specify it as num1/num2, where num1 and num2 are specified traditionally. This is useful for specifying a width/height ratio of 4/3, without having to write infinitely many digits. Where num is supposed to be a natural number, pamperspective does not allow this format.

Quadrilateral specification options

--upper_left_x=num

--ulx=num

This specifies the horizontal coordinate of the upper left vertex of the quadrilateral. The meaning of 'upper left' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--upper_left_y=num

--uly=num

This specifies the vertical coordinate of the upper left vertex of the quadrilateral. The meaning of 'upper left' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--upper_right_x=num

--urx=num

This specifies the horizontal coordinate of the upper right vertex of the quadrilateral. The meaning of 'upper right' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--upper_right_y=num

--ury=num

This specifies the vertical coordinate of the upper right vertex of the quadrilateral. The meaning of 'upper right' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--lower_left_x=num

--llx=num

This specifies the horizontal coordinate of the lower left vertex of the quadrilateral. The meaning of 'lower left' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--lower_left_y=num

--lly=num

This specifies the vertical coordinate of the lower left vertex of the quadrilateral. The meaning of 'lower left' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--lower_right_x=num

--lrx=num

This specifies the horizontal coordinate of the lower right vertex of the quadrilateral. The meaning of 'lower right' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--lower_right_y=num

--lry=num

This specifies the vertical coordinate of the lower right vertex of the quadrilateral. The meaning of 'lower right' is relative to the output image. The interpretation of num depends on the values for --input_system and --input_unit.

--input_system=system

--input_unit=unit

The input image consists of pixels, which are, from the point of view of a scanline renderer, solid squares. These options specify how the coordinates are interpreted:

system=lattice, unit=image: (0,0) refers to the upper left corner of the upper left pixel and (1,1) refers to the lower right corner of the lower right pixel.
system=lattice, unit=pixel: (0,0) refers to the upper left corner of the upper left pixel and (width,height) refers to the lower right corner of the lower right pixel. Here width and height are the width and height of the input image.
system=pixel, unit=image: (0,0) refers to the center of the upper left pixel and (1,1) refers to the center of the lower right pixel.
system=pixel, unit=pixel: (0,0) refers to the center of the upper left pixel and (width-1,height-1) refers to the center of the lower right pixel. Here width and height are the width and height of the input image.

The defaults are --input_system=lattice and --input_unit=pixel. Point-and-click front ends should use --input_system=pixel.

Frame Options

By default pamperspective outputs exactly the above parallelogram, sheared to a rectangle. With the following options, it is possible to make pamperspective output a larger or smaller portion, which we call the "visible part." We refer to the default rectangle as the "frame." The visible part is always a rectangle the axes of which are parallel to those of the frame.

The frame options are additive. All the parts of the image specified by either margin options, --include_frame, or --include (or their defaults) are in the visible part. The visible part is the smallest possible rectangle that contains the parts specified those three ways.

The visible part must have nonzero size. That means if you specify --frame_include=no (overriding the default), you'll need to specify other frame options in order to have something in the visible part.

[--margin=num]

This specifies an area surrounding the frame that is to be included in the visible part. The units of num are the width of the frame for the horizontal extensions and the height of the frame for vertical extensions.

For example, --margin=1 makes the visible part 9 times as large, because it makes the visible part extend one frame's worth to the left of the frame, one frame's worth to the right, one frame's worth above the frame, and one frame's worth below the frame, for a total of 3 frames' worth in both dimensions.

A negative value has an effect only if you specify --frame_include=no. The default is no margin.

The individual margin options below override this common margin setting.

[--top_margin=num]

[--left_margin=num]

[--right_margin=num]

[--bottom_margin=num]

These are like --margin, but they specify only one of the 4 sides. The default value for each is the value (or default) of --margin.

[--frame_include=bool]

Valid values for bool are:

yes
true
on: The frame itself is in the visible part.
no
false
off: The frame itself is not necessarily in the visible part (but it could be if other options cause it to be).

The default value is yes

--include=[x1,y1;x2,y2; ...]

The visible part is made large enough such that every point (x1,y1), (x2,y2), of the input image is visible. The meaning of x and y is determined by --input_system and --input_unit. You can specify any number of semicolon-delimited points, including zero.

If you're supplying these options via a Unix command shell, be sure to use proper quoting, because semicolon (;) is usually a shell control character.

The frame options were new in Netpbm 10.25 (October 2004).

Output Size Options

--width=width

--height=height

These specify the size of the output image in horizontal and vertical direction. The values are numbers of pixels, so only natural numbers are valid. These values override the default means to determine the output size.

--detail=num

If you do not specify --width, pamperspective determines the width of the output image such that moving num output pixels horizontally does not change the corresponding pixel coordinates of the input image by more than 1. pamperspective determines the height of the output image analogously. The default value is 1.

--proportion=prop

--ratio=ratio

Valid values for prop are:

free: In this case --ratio does not have any effect.
fixed: After the width and height are determined according to --detail, one of both will be increased, in order to obtain width/height=ratio.

The defaults are --proportion=free and --ratio=1.

Output Options

--output_system=spec

The output image consists of pixels, which are, from the point of view of a scanline renderer, solid squares. This option specifies how the four vertices of the quadrilateral correspond to the pixels of the output image. Valid values for spec are:

lattice: The upper left vertex corresponds to the upper left corner of the upper left pixel and The lower right vertex corresponds to the lower right corner of the lower right pixel.
pixel: The upper left vertex corresponds to the center of the upper left pixel and The lower right vertex corresponds to the center of the lower right pixel.

The default value is lattice. Point-and-click front ends should use pixel.

--interpolation=spec

Usually (centers of) output pixels do not exactly correspond to (centers of) input pixels. This option determines how the program will choose the new pixels. Valid values for spec are:

nearest: The output pixel will be identical to the nearest input pixel.
linear: The output pixel will be a bilinear interpolation of the four surrounding input pixels.

The default value is nearest.

HINTS

It might be tempting always to use the options --include 0,0;0,1;1,0;1,1 (assuming --input_system=lattice and --input_unit=image), so that no part of the input image is missing in the output. There are problems with that:

If the three dimensional plane defined by the quadrilateral has a visible horizon in the input image, then the above asks pamperspective to include points that cannot ever be part of the output.
If the horizon is not visible, but close to the border of the input image, this may result in very large output files. Consider a picture of a road. If you ask for a point close to the horizon to be included, then this point is far away from the viewer. The output will cover many kilometers of road, while --detail perhaps makes a pixel represent a square centimeter.

When working with large files pamperspective's memory usage might be an issue. In order to keep it small, you should minimize each of the following:

The vertical range that the top output line consumes in the input image;
The vertical range that the bottom output line consumes in the input image;
The vertical range from the topmost (with respect to the input image) quadrilateral point to the top (with respect to the output image) output line.

For this purpose you can use pamflip before and/or after pamperspective. Example: Instead of

pamperspective 10 0 100 50 0 20 95 100 infile > outfile

you can use


pamflip -rotate90 infile | 
   pamperspective 50 0 100 5 0 90 20 100 | 
   pamflip -rotate270 > outfile

HISTORY

Mark Weyer wrote pamperspective in March 2004.

It was new in Netpbm 10.22 (April 2004).

AUTHOR

This documentation was written by Mark Weyer. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU General Public License, Version 2 or any later version published by the Free Software Foundation.

SYNOPSIS
DESCRIPTION
OPTIONS
HINTS
SEE ALSO
HISTORY
AUTHOR