This is an in-house project that was carried out to develop expertise on JPEG and JPEG 2000. We developed a baseline sequential JPEG decoder software conforming to ISO/IEC 10918-1: 1994 (E) standard. It takes a compressed JPEG file in binary format as input and generates a windows bitmap file as output. It uses 16-bit fixed point computations. The reference JPEG implementation uses floating point computations and is as such not suitable for embedded devices. The floating point calculations were converted to 16-bit fixed point calculations without significant loss of precision.

The jpeg decoder has been ported to ARM7TDMI.

Salient features of the decoder are:

• Supports valid marker sequences for baseline sequential as per the standard ISO/IEC - 10918-1 : 1994-15-12.
• Supports the commonly used YCrCb color space and grey scale modes in input. Produces output in windows bitmap format in 24-bit TRUE color or grey scale mode.
• Error checking for invalid parameters, huffman tables and dequantization tables.
• Implemented using 16-bit fixed point data.
• Critical and complex mathematical/ trigonometric computations achieved using table look-up.