IPL to PFM Converter

PFM (Portable Float Map) is a floating-point raster image format devised by Paul Debevec around 2001, designed to store high-dynamic-range image data with the simplicity of the Netpbm family of formats. PFM extends the PBM/PGM/PPM philosophy — minimal header, raw data, no compression — to 32-bit IEEE floating-point samples, providing direct access to HDR pixel values without the encoding overhead of formats like OpenEXR or the limited range of Radiance HDR's RGBE encoding. The file structure is deliberately minimal: a two-character magic number ('Pf' for grayscale, 'PF' for color), width and height on the next line, a scale/endianness indicator (negative for little-endian, positive for big-endian, with magnitude indicating scale factor), and then the raw 32-bit float data for each pixel. PFM files store one float per pixel for grayscale or three floats (RGB) per pixel for color, with no compression, alpha channel, or metadata support. The format emerged from the HDR imaging research community where Debevec's work on image-based lighting and light stage capture required a simple, unambiguous way to store linear floating-point radiance values that could be easily exchanged between research tools. One advantage is absolute simplicity for HDR data: PFM can be read and written in a few lines of code in any language that supports IEEE floats, with no library dependencies — ideal for research prototyping and quick data exchange between custom tools. The format's widespread adoption in the computer vision and computational photography research community is another practical strength — optical flow benchmarks (Middlebury), depth estimation datasets, and radiance field captures commonly use PFM. The format is supported by ImageMagick, OpenCV, HDR Shop, and Luminance HDR.