.PFM File Format

Introduction

The Portable Float Map is the simplest way to store an image of floating point values. It follows the idea of the Netpbm formats: a few lines of text, then the raw pixel values with no compression. Each sample is a 32-bit float, so the file can hold any brightness, not just 0 to 255.

It is used for high dynamic range photographs and renders, and in computer vision research for depth and disparity maps. The files are large, but any program can read them with a few lines of code.

Structure

The header is three lines of text. The first is PF for a colour image with three channels, or Pf for one greyscale channel. The second gives the width and height. The third is a scale factor whose sign gives the byte order: a negative number means little-endian floats, a positive one big-endian.

After the header come the pixels as 32-bit floats, with no padding. Rows are stored from the bottom of the image to the top, the opposite of most Netpbm formats.

Also used for

• PostScript Type 1 font metrics – Adobe’s PostScript font: cubic outlines in an encrypted PostScript program.

Links