The standard defined by JPEG is useful in a broad range of applications. Because each application has different compression requirements, several processes for compression and decompression are specified within the JPEG standard. The processes fall into three general categories: the baseline sequential process, the extended DCT-based processes, and the lossless process. All JPEG coders and decoders must support the baseline sequential process. All other processes are optional extensions that can be useful in specific applications. For detailed information on each process, refer to the ISO Committee document, ISO/IEC CD 10918-1.

Lossless JPEG uses DCPM and lossy JPEG uses DCT transform. Compression can be lossless at rates of about 2:1 or lossy, with higher compression. Nearly imperceptible levels of distortion are introduced at about 20:1

The baseline sequential process is based on the discrete cosine transform (DCT) followed by variable-word-length coding (Huffman coding). This process provides substantial compression (up to 100:1) while maintaining a high degree of visual fidelity in the reconstructed image. DCT-based processes, however, are lossy processes. The reconstructed images are not byte-for-byte equivalent to the source images. Further, the level of loss in the image varies with the compression ratio. Typically, the baseline sequential process can compress image data to about one bit per pixel, or less, with good visual quality in the reconstructed image. For example, a 24-bit RGB color image can be compressed to one bit per pixel (less than five percent of the original size), and the reconstructed image will be nearly indistinguishable from the original.