image.h File Reference

Image loading, preprocessing, and normalisation. More...

#include <stddef.h>
#include <stdint.h>

Include dependency graph for image.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	Image
	Heap-allocated RGBA pixel buffer (row-major, 4 bytes per pixel). More...
struct	PreprocessParams
	Parameters for the full preprocessing pipeline. More...

Functions
Image *	image_load_png (const char *path)
	Load a PNG file into an Image.
void	image_free (Image *img)
	Free an Image allocated by image_load_png().
void	image_to_grayscale (Image *img)
	Convert an Image to grayscale in-place.
void	image_binarize (Image *img)
	Binarise a grayscale Image in-place using a global threshold.
Image *	image_resize (const Image *img, int w, int h)
	Resize an Image to the given dimensions (nearest-neighbour).
Image *	image_rotate (const Image *img, float angle_deg)
	Rotate an Image by angle_deg degrees clockwise.
void	image_to_float (const Image *img, float *out)
	Convert an Image to a flat normalised float array.
int	image_load_normalised (const char *path, const PreprocessParams *p, float *out, int out_h, int out_w)
	Full pipeline: load PNG → preprocess → resize → float array.
int	image_save_png (const Image *img, const char *path)
	Save an Image as a PNG file.

Detailed Description

Image loading, preprocessing, and normalisation.

Uses libpng for PNG loading (no SDL2_image required). SDL2 is used only for rotation rendering.

Pipeline applied before CNN inference: load PNG → grayscale → binarize → (optional rotate) → resize 28×28 → float[784]

Convention: black ink = 1.0, white background = 0.0.

Function Documentation

image_binarize()

void image_binarize

(

Image *

img

)

Binarise a grayscale Image in-place using a global threshold.

The threshold is the mean pixel luminance of the image. Pixels above the threshold → 255 (white/background). Pixels at or below → 0 (black/ink).

Parameters

img	Grayscale Image (R channel used as luminance).

Here is the caller graph for this function:

image_free()

void image_free

(

Image *

img

)

Free an Image allocated by image_load_png().

Parameters

img	Image to free. No-op if NULL.

Here is the caller graph for this function:

image_load_normalised()

int image_load_normalised	(	const char *	path,
		const PreprocessParams *	p,
		float *	out,
		int	out_h,
		int	out_w )

Full pipeline: load PNG → preprocess → resize → float array.

Parameters

path	Path to the source PNG file.
p	Preprocessing parameters (may be NULL for defaults).
out	Output buffer of size out_h * out_w floats.
out_h	Target height (e.g. 28 for CNN input).
out_w	Target width (e.g. 28 for CNN input).

Returns

0 on success, -1 on error.

Here is the call graph for this function:

Here is the caller graph for this function:

image_load_png()

Image * image_load_png

(

const char *

path

)

Load a PNG file into an Image.

Parameters

path	Path to the PNG file.

Returns

Heap-allocated Image on success, or NULL on error. Free with image_free().

Here is the call graph for this function:

Here is the caller graph for this function:

image_resize()

Image * image_resize	(	const Image *	img,
		int	w,
		int	h )

Resize an Image to the given dimensions (nearest-neighbour).

Returns a new Image; the original is not modified.

Parameters

img	Source Image.
w	Target width in pixels.
h	Target height in pixels.

Returns

New Image on success, NULL on allocation failure. Free with image_free().

Here is the call graph for this function:

Here is the caller graph for this function:

image_rotate()

Image * image_rotate	(	const Image *	img,
		float	angle_deg )

Rotate an Image by angle_deg degrees clockwise.

Returns a new Image large enough to contain the full rotated content. Background pixels are set to white (255, 255, 255, 255).

Parameters

img	Source Image.
angle_deg	Rotation angle in degrees (clockwise).

Returns

New rotated Image, or NULL on failure. Free with image_free().

Here is the call graph for this function:

Here is the caller graph for this function:

image_save_png()

int image_save_png	(	const Image *	img,
		const char *	path )

Save an Image as a PNG file.

Parameters

img	Image to save.
path	Destination file path.

Returns

0 on success, -1 on error.

image_to_float()

void image_to_float	(	const Image *	img,
		float *	out )

Convert an Image to a flat normalised float array.

Reads the R channel (expected to hold grayscale luminance). Ink (R=0) → 1.0f, background (R=255) → 0.0f.

Parameters

img	Source Image (must be grayscale).
out	Output buffer of size img->width * img->height floats.

Here is the caller graph for this function:

image_to_grayscale()

void image_to_grayscale

(

Image *

img

)

Convert an Image to grayscale in-place.

Uses ITU-R BT.601 luminance weights: L = 0.299·R + 0.587·G + 0.114·B

After this call the R, G, and B channels all hold the luminance value.

Parameters

img	Image to convert.

Here is the caller graph for this function: