# PNMImage¶

class PNMImage

Bases: PNMImageHeader

The name of this class derives from the fact that we originally implemented it as a layer on top of the “pnm library”, based on netpbm, which was built to implement pbm, pgm, and pbm files, and is the underlying support of a number of public-domain image file converters. Nowadays we are no longer derived directly from the pnm library, mainly to allow support of C++ iostreams instead of the C stdio FILE interface.

Conceptually, a PNMImage is a two-dimensional array of xels, which are the PNM-defined generic pixel type. Each xel may have a red, green, and blue component, or (if the image is grayscale) a gray component. The image may be read in, the individual xels manipulated, and written out again, or a black image may be constructed from scratch.

A PNMImage has a color space and a maxval, the combination of which defines how a floating-point linear color value is encoded as an integer value in memory. The functions ending in _val operate on encoded colors, whereas the regular ones work with linear floating-point values. All operations are color space correct unless otherwise specified.

The image is of size XSize() by YSize() xels, numbered from top to bottom, left to right, beginning at zero.

Files can be specified by filename, or by an iostream pointer. The filename “-” refers to stdin or stdout.

This class is not inherently thread-safe; use it from a single thread or protect access using a mutex.

Inheritance diagram

class CRow

Provides an accessor for reading the contents of one row of the image in- place.

CRow(PNMImage::CRow const&) = default
xelval get_alpha_val(int x) const

Fetch the alpha value at the given column in the row.

xel get_xel_val(int x) const

Fetch the pixel at the given column in the row.

std::size_t size(void) const

Get the number of pixels in the row.

class Row

Provides an accessor for reading or writing the contents of one row of the image in-place.

Row(PNMImage::Row const&) = default
xelval get_alpha_val(int x) const

Fetch the alpha value at the given column in the row.

xel &get_xel_val(int x)

Fetch the pixel at the given column in the row.

void set_alpha_val(int x, xelval v)

Set the alpha value at the given column in the row.

void set_xel_val(int x, xel const &v)

Set the pixel at the given column in the row.

std::size_t size(void) const

Get the number of pixels in the row.

PNMImage(void)
explicit PNMImage(Filename const &filename, PNMFileType *type = nullptr)
explicit PNMImage(int x_size, int y_size, int num_channels = 3, xelval maxval = 255, PNMFileType *type = nullptr, ColorSpace color_space = ::CS_linear)
PNMImage(PNMImage const &copy)
void add_alpha(void)

Adds an alpha channel to the image, if it does not already have one. The alpha channel is initialized to zeros.

void add_sub_image(PNMImage const &copy, int xto, int yto, int xfrom = 0, int yfrom = 0, int x_size = -1, int y_size = -1, float pixel_scale = 1.0)

Behaves like copy_sub_image(), except the copy pixels are added to the pixels of the destination, after scaling by the specified pixel_scale. Unlike blend_sub_image(), the alpha channel is not treated specially.

void alpha_fill(float alpha = 0.0)

Sets the entire alpha channel to the given level.

void alpha_fill_val(xelval alpha = 0)

Sets the entire alpha channel to the given level.

void apply_exponent(float gray_exponent)
void apply_exponent(float gray_exponent, float alpha_exponent)
void apply_exponent(float red_exponent, float green_exponent, float blue_exponent)
void apply_exponent(float red_exponent, float green_exponent, float blue_exponent, float alpha_exponent)

Adjusts each channel of the image by raising the corresponding component value to the indicated exponent, such that L’ = L ^ exponent.

Adjusts each channel of the image by raising the corresponding component value to the indicated exponent, such that L’ = L ^ exponent.

Adjusts each channel of the image by raising the corresponding component value to the indicated exponent, such that L’ = L ^ exponent. For a grayscale image, the blue_exponent value is used for the grayscale value, and red_exponent and green_exponent are unused.

Adjusts each channel of the image by raising the corresponding component value to the indicated exponent, such that L’ = L ^ exponent. For a grayscale image, the blue_exponent value is used for the grayscale value, and red_exponent and green_exponent are unused.

void blend(int x, int y, LRGBColorf const &val, float alpha)
void blend(int x, int y, float r, float g, float b, float alpha)

Smoothly blends the indicated pixel value in with whatever was already in the image, based on the given alpha value. An alpha of 1.0 is fully opaque and completely replaces whatever was there previously; alpha of 0.0 is fully transparent and does nothing.

void blend_sub_image(PNMImage const &copy, int xto, int yto, int xfrom = 0, int yfrom = 0, int x_size = -1, int y_size = -1, float pixel_scale = 1.0)

Behaves like copy_sub_image(), except the alpha channel of the copy is used to blend the copy into the destination image, instead of overwriting pixels unconditionally.

If pixel_scale is not 1.0, it specifies an amount to scale each alpha value of the source image before applying it to the target image.

If pixel_scale is 1.0 and the copy has no alpha channel, this degenerates into copy_sub_image().

void box_filter(float radius = 1.0)

This flavor of box_filter() will apply the filter over the entire image without resizing or copying; the effect is that of a blur operation.

void box_filter_from(float radius, PNMImage const &copy)

Makes a resized copy of the indicated image into this one using the indicated filter. The image to be copied is squashed and stretched to match the dimensions of the current image, applying the appropriate filter to perform the stretching.

xelval clamp_val(int input_value) const

A handy function to clamp values to [0..get_maxval()].

void clear(void)
void clear(int x_size, int y_size, int num_channels = 3, xelval maxval = 255, PNMFileType *type = nullptr, ColorSpace color_space = ::CS_linear)

Frees all memory allocated for the image, and clears all its parameters (size, color, type, etc).

This flavor of clear() reinitializes the image to an empty (black) image with the given dimensions.

void clear_read_size(void)

Undoes the effect of a previous call to set_read_size().

void copy_channel(PNMImage const &copy, int src_channel, int dest_channel)
void copy_channel(PNMImage const &copy, int xto, int yto, int cto, int xfrom = 0, int yfrom = 0, int cfrom = 0, int x_size = -1, int y_size = -1)

Copies a channel from one image into another. Images must be the same size

Copies just a single channel from the source image into a single channel of this image, leaving the remaining channels alone.

void copy_channel_bits(PNMImage const &copy, int src_channel, int dest_channel, xelval src_mask, int right_shift)

Copies some subset of the bits of the specified channel from one image into some subset of the bits of the specified channel in another image. Images must be the same size.

If right_shift is negative, it means a left shift.

void copy_from(PNMImage const &copy)

Makes this image become a copy of the other image.

void copy_header_from(PNMImageHeader const &header)

Copies just the header information into this image. This will blow away any image data stored in the image. The new image data will be allocated, but left unitialized.

void copy_sub_image(PNMImage const &copy, int xto, int yto, int xfrom = 0, int yfrom = 0, int x_size = -1, int y_size = -1)

Copies a rectangular area of another image into a rectangular area of this image. Both images must already have been initialized. The upper-left corner of the region in both images is specified, and the size of the area; if the size is omitted, it defaults to the entire other image, or the largest piece that will fit.

void darken_sub_image(PNMImage const &copy, int xto, int yto, int xfrom = 0, int yfrom = 0, int x_size = -1, int y_size = -1, float pixel_scale = 1.0)

Behaves like copy_sub_image(), but the resulting color will be the darker of the source and destination colors at each pixel (and at each R, G, B, A component value).

If pixel_scale is not 1.0, it specifies an amount to scale each pixel value of the source image before applying it to the target image. The scale is applied with the center at 1.0: scaling the pixel value smaller brings it closer to 1.0.

void do_fill_distance(int xi, int yi, int d)

Recursively fills in the minimum distance measured from a certain set of points into the gray channel.

void expand_border(int left, int right, int bottom, int top, LColorf const &color)

Expands the image by the indicated number of pixels on each edge. The new pixels are set to the indicated color.

If any of the values is negative, this actually crops the image.

void fill(float red, float green, float blue)
void fill(float gray = 0.0)

Sets the entire image (except the alpha channel) to the given color.

Sets the entire image (except the alpha channel) to the given grayscale level.

void fill_distance_inside(PNMImage const &mask, float threshold, int radius, bool shrink_from_border)

Replaces this image with a grayscale image whose gray channel represents the linear Manhattan distance from the nearest dark pixel in the given mask image, up to the specified radius value (which also becomes the new maxval). radius may range from 0 to maxmaxval; smaller values will compute faster. A dark pixel is defined as one whose pixel value is < threshold.

If shrink_from_border is true, then the mask image is considered to be surrounded by a border of dark pixels; otherwise, the border isn’t considered.

This can be used, in conjunction with threshold, to shrink a mask image inwards by a certain number of pixels.

The mask image may be the same image as this one, in which case it is destructively modified by this process.

void fill_distance_outside(PNMImage const &mask, float threshold, int radius)

Replaces this image with a grayscale image whose gray channel represents the linear Manhattan distance from the nearest white pixel in the given mask image, up to the specified radius value (which also becomes the new maxval). radius may range from 0 to maxmaxval; smaller values will compute faster. A white pixel is defined as one whose pixel value is >= threshold.

This can be used, in conjunction with threshold, to grow a mask image outwards by a certain number of pixels.

The mask image may be the same image as this one, in which case it is destructively modified by this process.

void fill_val(xelval red, xelval green, xelval blue)
void fill_val(xelval gray = 0)

Sets the entire image (except the alpha channel) to the given grayscale level.

Sets the entire image (except the alpha channel) to the given color.

void flip(bool flip_x, bool flip_y, bool transpose)

Reverses, transposes, and/or rotates the image in-place according to the specified parameters. If flip_x is true, the x axis is reversed; if flip_y is true, the y axis is reversed. Then, if transpose is true, the x and y axes are exchanged. These parameters can be used to select any combination of 90-degree or 180-degree rotations and flips.

float from_alpha_val(xelval input_value) const

A handy function to scale alpha values from [0..get_maxval()] to [0..1].

LRGBColorf from_val(xel const &input_value) const
float from_val(xelval input_value) const

A handy function to scale non-alpha values from [0..get_maxval()] to [0..1]. Do not use this for alpha values, see from_alpha_val().

void gamma_correct(float from_gamma, float to_gamma)

Assuming the image was constructed with a gamma curve of from_gamma in the RGB channels, converts it to an image with a gamma curve of to_gamma in the RGB channels. Does not affect the alpha channel.

void gamma_correct_alpha(float from_gamma, float to_gamma)

Assuming the image was constructed with a gamma curve of from_gamma in the alpha channel, converts it to an image with a gamma curve of to_gamma in the alpha channel. Does not affect the RGB channels.

void gaussian_filter(float radius = 1.0)

This flavor of gaussian_filter() will apply the filter over the entire image without resizing or copying; the effect is that of a blur operation.

void gaussian_filter_from(float radius, PNMImage const &copy)

Makes a resized copy of the indicated image into this one using the indicated filter. The image to be copied is squashed and stretched to match the dimensions of the current image, applying the appropriate filter to perform the stretching.

float get_alpha(int x, int y) const

Returns the alpha component color at the indicated pixel. It is an error to call this unless has_alpha() is true. The value returned is a float in the range 0..1.

xelval get_alpha_val(int x, int y) const

Returns the alpha component color at the indicated pixel. It is an error to call this unless has_alpha() is true. The value returned is in the range 0..maxval and always linear.

float get_average_gray(void) const

Returns the average grayscale component of all of the pixels in the image.

LRGBColorf get_average_xel(void) const

Returns the average color of all of the pixels in the image.

LColorf get_average_xel_a(void) const

Returns the average color of all of the pixels in the image, including the alpha channel.

float get_blue(int x, int y) const

Returns the blue component color at the indicated pixel. The value returned is a linearized float in the range 0..1.

xelval get_blue_val(int x, int y) const

Returns the blue component color at the indicated pixel. The value returned is in the range 0..maxval and encoded in the configured color space.

float get_bright(int x, int y) const
float get_bright(int x, int y, float rc, float gc, float bc) const
float get_bright(int x, int y, float rc, float gc, float bc, float ac) const

Returns the linear brightness of the given xel, as a linearized float in the range 0..1. This flavor of get_bright() returns the correct grayscale brightness level for both full-color and grayscale images.

This flavor of get_bright() works correctly only for color images. It returns a single brightness value for the RGB color at the indicated pixel, based on the supplied weights for each component.

This flavor of get_bright() works correctly only for four-channel images. It returns a single brightness value for the RGBA color at the indicated pixel, based on the supplied weights for each component.

float get_channel(int x, int y, int channel) const

Returns the nth component color at the indicated pixel. The channel index should be in the range 0..(get_num_channels()-1). The channels are ordered B, G, R, A. This is slightly less optimal than accessing the component values directly by named methods. The value returned is a float in the range 0..1.

xelval get_channel_val(int x, int y, int channel) const

Returns the nth component color at the indicated pixel. The channel index should be in the range 0..(get_num_channels()-1). The channels are ordered B, G, R, A. This is slightly less optimal than accessing the component values directly by named methods. The value returned is in the range 0..maxval.

ColorSpace get_color_space(void) const

Returns the color space in which the image is encoded.

float get_gray(int x, int y) const

Returns the gray component color at the indicated pixel. This only has a meaningful value for grayscale images; for other image types, this returns the value of the blue channel only. However, also see the get_bright() function. The value returned is a linearized float in the range 0..1.

xelval get_gray_val(int x, int y) const

Returns the gray component color at the indicated pixel. This only has a meaningful value for grayscale images; for other image types, this returns the value of the blue channel only. However, also see the get_bright() function. The value returned is in the range 0..maxval and encoded in the configured color space.

float get_green(int x, int y) const

Returns the green component color at the indicated pixel. The value returned is a linearized float in the range 0..1.

xelval get_green_val(int x, int y) const

Returns the green component color at the indicated pixel. The value returned is in the range 0..maxval and encoded in the configured color space.

PNMImageHeader::PixelSpec get_pixel(int x, int y) const

Returns the (r, g, b, a) pixel value at the indicated pixel, using a PixelSpec object.

int get_read_x_size(void) const

Returns the requested x_size of the image if set_read_size() has been called, or the image x_size otherwise (if it is known).

int get_read_y_size(void) const

Returns the requested y_size of the image if set_read_size() has been called, or the image y_size otherwise (if it is known).

float get_red(int x, int y) const

Returns the red component color at the indicated pixel. The value returned is a linearized float in the range 0..1.

xelval get_red_val(int x, int y) const

Returns the red component color at the indicated pixel. The value returned is in the range 0..maxval and encoded in the configured color space.

LRGBColorf get_xel(int x, int y) const

Returns the RGB color at the indicated pixel. Each component is a linearized float in the range 0..1.

LColorf get_xel_a(int x, int y) const

Returns the RGBA color at the indicated pixel. Each component is a linearized float in the range 0..1.

xel &get_xel_val(int x, int y)
xel get_xel_val(int x, int y) const

Returns the RGB color at the indicated pixel. Each component is in the range 0..maxval.

bool has_read_size(void) const

Returns true if set_read_size() has been called.

void indirect_1d_lookup(PNMImage const &index_image, int channel, PNMImage const &pixel_values)

index_image is a WxH grayscale image, while pixel_values is an Nx1 color (or grayscale) image. Typically pixel_values will be a 256x1 image.

Fills the PNMImage with a new image the same width and height as index_image, with the same number of channels as pixel_values.

Each pixel of the new image is computed with the formula:

new_image(x, y) = pixel_values(index_image(x, y)[channel], 0)

At present, no interpolation is performed; the nearest value in pixel_values is discovered. This may change in the future.

bool is_valid(void) const

Returns true if the image has been read in or correctly initialized with a height and width. If this returns false, virtually all member functions except clear() and read() are invalid function calls.

void lighten_sub_image(PNMImage const &copy, int xto, int yto, int xfrom = 0, int yfrom = 0, int x_size = -1, int y_size = -1, float pixel_scale = 1.0)

Behaves like copy_sub_image(), but the resulting color will be the lighter of the source and destination colors at each pixel (and at each R, G, B, A component value).

If pixel_scale is not 1.0, it specifies an amount to scale each pixel value of the source image before applying it to the target image.

void make_grayscale(void)
void make_grayscale(float rc, float gc, float bc)

Converts the image from RGB to grayscale. Any alpha channel, if present, is left undisturbed.

Converts the image from RGB to grayscale. Any alpha channel, if present, is left undisturbed. The optional rc, gc, bc values represent the relative weights to apply to each channel to convert it to grayscale.

void make_histogram(PNMImageHeader::Histogram &hist)

Computes a histogram of the colors used in the image.

void make_rgb(void)

Converts the image from grayscale to RGB. Any alpha channel, if present, is left undisturbed.

void mult_sub_image(PNMImage const &copy, int xto, int yto, int xfrom = 0, int yfrom = 0, int x_size = -1, int y_size = -1, float pixel_scale = 1.0)

Behaves like copy_sub_image(), except the copy pixels are multiplied to the pixels of the destination, after scaling by the specified pixel_scale. Unlike blend_sub_image(), the alpha channel is not treated specially.

void perlin_noise_fill(float sx, float sy, int table_size = 256, unsigned long int seed = 0)
void perlin_noise_fill(StackedPerlinNoise2 &perlin)

Fills the image with a grayscale perlin noise pattern based on the indicated parameters. Uses set_xel() to set the grayscale values. The sx and sy parameters are in multiples of the size of this image. See also the PerlinNoise2 class in mathutil.

Variant of perlin_noise_fill that uses an existing StackedPerlinNoise2 object.

void premultiply_alpha(void)

Converts an image in-place to its “premultiplied” form, where, for every pixel in the image, the red, green, and blue components are multiplied by that pixel’s alpha value.

This does not modify any alpha values.

void quantize(std::size_t max_colors)

Reduces the number of unique colors in the image to (at most) the given count. Fewer colors than requested may be left in the image after this operation, but never more.

At present, this is only supported on images without an alpha channel.

New in version 1.10.5.

void quick_filter_from(PNMImage const &copy, int xborder = 0, int yborder = 0)

Resizes from the given image, with a fixed radius of 0.5. This is a very specialized and simple algorithm that doesn’t handle dropping below the Nyquist rate very well, but is quite a bit faster than the more general box_filter(), above. If borders are specified, they will further restrict the size of the resulting image. There’s no point in using quick_box_filter() on a single image.

bool read(Filename const &filename, PNMFileType *type = nullptr, bool report_unknown_type = true)
bool read(std::istream &data, std::string const &filename = string(), PNMFileType *type = nullptr, bool report_unknown_type = true)
bool read(PNMReader *reader)

Reads the indicated image filename. If type is non-NULL, it is a suggestion for the type of file it is. Returns true if successful, false on error.

Reads the image data from the indicated stream.

The filename is advisory only, and may be used to suggest a type if it has a known extension.

If type is non-NULL, it is a suggestion for the type of file it is (and a non-NULL type will override any magic number test or filename extension lookup).

Returns true if successful, false on error.

This flavor of read() uses an already-existing PNMReader to read the image file. You can get a reader via the PNMImageHeader::make_reader() methods. This is a good way to examine the header of a file (for instance, to determine its size) before actually reading the entire image.

The PNMReader is always deleted upon completion, whether successful or not.

void remix_channels(LMatrix4 const &conv)

Transforms every pixel using the operation (Ro,Go,Bo) = conv.xform_point(Ri,Gi,Bi); Input must be a color image.

void remove_alpha(void)

Removes the image’s alpha channel, if it exists.

void render_spot(LColorf const &fg, LColorf const &bg, float min_radius, float max_radius)

Renders a solid-color circle, with a fuzzy edge, into the center of the PNMImage. If the PNMImage is non-square, this actually renders an ellipse.

The min_radius and max_radius are in the scale 0..1, where 1.0 means the full width of the image. If min_radius == max_radius, the edge is sharp (but still antialiased); otherwise, the pixels between min_radius and max_radius are smoothly blended between fg and bg colors.

void rescale(float min_val, float max_val)

Rescales the RGB channel values so that any values in the original image between min_val and max_val are expanded to the range 0 .. 1. Values below min_val are set to 0, and values above max_val are set to 1. Does not affect the alpha channel, if any.

void reverse_rows(void)

Performs an in-place reversal of the row (y) data.

void set_alpha(int x, int y, float a)

Sets the alpha component color only at the indicated pixel. It is an error to call this unless has_alpha() is true. The value given should be in the range 0..1.

void set_alpha_val(int x, int y, xelval a)

Sets the alpha component color only at the indicated pixel. It is an error to call this unless has_alpha() is true. The value given should be in the range 0..maxval.

This value is always linearly encoded, even if the image is set to the sRGB color space.

void set_blue(int x, int y, float b)

Sets the blue component color only at the indicated pixel. The value given should be a linearized float in the range 0..1.

void set_blue_val(int x, int y, xelval b)

Sets the blue component color only at the indicated pixel. The value given should be in the range 0..maxval, encoded in the configured color space. See set_blue() if you instead have a linearized and normalized floating-point value.

void set_channel(int x, int y, int channel, float value)

Sets the nth component color at the indicated pixel. The channel index should be in the range 0..(get_num_channels()-1). The channels are ordered B, G, R, A. This is slightly less optimal than setting the component values directly by named methods. The value given should be a float in the range 0..1.

void set_channel_val(int x, int y, int channel, xelval value)

Sets the nth component color at the indicated pixel. The channel index should be in the range 0..(get_num_channels()-1). The channels are ordered B, G, R, A. This is slightly less optimal than setting the component values directly by named methods. The value given should be in the range 0..maxval.

void set_color_space(ColorSpace color_space)

Converts the colors in the image to the indicated color space. This may be a lossy operation, in particular when going from sRGB to linear. The alpha channel remains untouched.

Note that, because functions like get_xel() and set_xel() work on linearized floating-point values, this conversion won’t affect those values (aside from some minor discrepancies due to storage precision). It does affect the values used by get_xel_val() and set_xel_val(), though, since those operate on encoded colors.

Some color spaces, particularly scRGB, may enforce the use of a particular maxval setting.

void set_color_type(PNMImageHeader::ColorType color_type)

Translates the image to or from grayscale, color, or four-color mode. Grayscale images are converted to full-color images with R, G, B set to the original gray level; color images are converted to grayscale according to the value of Bright(). The alpha channel, if added, is initialized to zero.

void set_gray(int x, int y, float gray)

Sets the gray component color at the indicated pixel. This is only meaningful for grayscale images; for other image types, this simply sets the blue component color. However, also see set_xel(), which can set all the component colors to the same grayscale level, and hence works correctly both for grayscale and color images. The value given should be a linearized float in the range 0..1.

void set_gray_val(int x, int y, xelval gray)

Sets the gray component color at the indicated pixel. This is only meaningful for grayscale images; for other image types, this simply sets the blue component color. However, also see set_xel_val(), which can set all the component colors to the same grayscale level, and hence works correctly both for grayscale and color images. The value given should be in the range 0..maxval, encoded in the configured color space. See set_gray() if you instead have a linearized normalized floating-point value.

void set_green(int x, int y, float g)

Sets the green component color only at the indicated pixel. The value given should be a linearized float in the range 0..1.

void set_green_val(int x, int y, xelval g)

Sets the green component color only at the indicated pixel. The value given should be in the range 0..maxval, encoded in the configured color space. See set_green() if you instead have a linearized and normalized floating-point value.

void set_maxval(xelval maxval)

Rescales the image to the indicated maxval.

void set_num_channels(int num_channels)

Changes the number of channels associated with the image. The new number of channels must be an integer in the range 1 through 4, inclusive. This will allocate and/or deallocate memory as necessary to accommodate; see set_color_type().

void set_pixel(int x, int y, PNMImageHeader::PixelSpec const &pixel)

Sets the (r, g, b, a) pixel value at the indicated pixel, using a PixelSpec object.

void set_read_size(int x_size, int y_size)

Specifies the size to we’d like to scale the image upon reading it. This will affect the next call to read(). This is usually used to reduce the image size, e.g. for a thumbnail.

If the file type reader supports it (e.g. JPEG), then this will scale the image during the read operation, consequently reducing memory and CPU utilization. If the file type reader does not support it, this will load the image normally, and them perform a linear scale after it has been loaded.

void set_red(int x, int y, float r)

Sets the red component color only at the indicated pixel. The value given should be a linearized float in the range 0..1.

void set_red_val(int x, int y, xelval r)

Sets the red component color only at the indicated pixel. The value given should be in the range 0..maxval, encoded in the configured color space. See set_red() if you instead have a linearized and normalized floating-point value.

void set_xel(int x, int y, LRGBColorf const &value)
void set_xel(int x, int y, float r, float g, float b)
void set_xel(int x, int y, float gray)

Changes the RGB color at the indicated pixel. Each component is a linearized float in the range 0..1.

Changes the RGB color at the indicated pixel. Each component is a linearized float in the range 0..1.

Changes all three color components at the indicated pixel to the same value. The value is a linearized float in the range 0..1.

void set_xel_a(int x, int y, LColorf const &value)
void set_xel_a(int x, int y, float r, float g, float b, float a)

Changes the RGBA color at the indicated pixel. Each component is a linearized float in the range 0..1.

void set_xel_val(int x, int y, xel const &value)
void set_xel_val(int x, int y, xelval r, xelval g, xelval b)
void set_xel_val(int x, int y, xelval gray)

Changes the RGB color at the indicated pixel. Each component is in the range 0..maxval, encoded in the configured color space. See set_xel() if you instead have a linearized and normalized floating-point value.

Changes the RGB color at the indicated pixel. Each component is in the range 0..maxval, encoded in the configured color space. See set_xel() if you instead have a linearized and normalized floating-point value.

Changes all three color components at the indicated pixel to the same value. The value is in the range component is in the range 0..maxval, encoded in the configured color space. See set_xel() if you instead have a linearized and normalized floating-point value.

void take_from(PNMImage &orig)

Move the contents of the other image into this one, and empty the other image.

void threshold(PNMImage const &select_image, int channel, float threshold, PNMImage const &lt, PNMImage const &ge)

Selectively copies each pixel from either one source or another source, depending on the pixel value of the indicated channel of select_image.

For each pixel (x, y):

s = select_image.get_channel(x, y, channel). Set this image’s (x, y) to:

lt.get_xel(x, y) if s < threshold, or

ge.get_xel(x, y) if s >= threshold

Any of select_image, lt, or ge may be the same PNMImge object as this image, or the same as each other; or they may all be different. All images must be the same size. As a special case, lt and ge may both be 1x1 images instead of the source image size.

xelval to_alpha_val(float input_value) const

A handy function to scale alpha values from [0..1] to [0..get_maxval()].

xel to_val(LRGBColorf const &input_value) const
xelval to_val(float input_value) const

A handy function to scale non-alpha values from [0..1] to [0..get_maxval()]. Do not use this for alpha values, see to_alpha_val().

void unfiltered_stretch_from(PNMImage const &copy)

Resizes from the indicated image into this one by performing a nearest- point sample.

void unpremultiply_alpha(void)

Converts an image in-place to its “straight alpha” form (presumably from a “premultiplied” form), where, for every pixel in the image, the red, green, and blue components are divided by that pixel’s alpha value.

This does not modify any alpha values.

bool write(Filename const &filename, PNMFileType *type = nullptr) const
bool write(std::ostream &data, std::string const &filename = string(), PNMFileType *type = nullptr) const
bool write(PNMWriter *writer) const

Writes the image to the indicated filename. If type is non-NULL, it is a suggestion for the type of image file to write.

Writes the image to the indicated ostream.

The filename is advisory only, and may be used suggest a type if it has a known extension.

If type is non-NULL, it is a suggestion for the type of image file to write.

This flavor of write() uses an already-existing PNMWriter to write the image file. You can get a writer via the PNMImageHeader::make_writer() methods.

The PNMWriter is always deleted upon completion, whether successful or not.