Abstract: A method and apparatus to compress raw color-filter array (CFA) image data in a mathematically lossless manner using spatial processing techniques. The raw CFA data is transformed using a prediction algorithm and then partitioned into blocks having N rows and M columns. The blocks are clustered into groups using a clustering algorithm. Next, compression is performed on blocks within a cluster before the compressed blocks are further compressed using arithmetic coding. In some implementations, alphabet creation and correlation is performed on the CFA data before prediction transformation, and color-plane correlation is performed after prediction transformation but before partitioning into blocks. This spatial compression method can be performed efficiently on camera to compress data files, reducing the storage and data flow requirements of the camera.

Abstract: A method and apparatus to compress raw color-filter array (CFA) image data in a mathematically lossless manner using spatial processing techniques. The raw CFA data is transformed using a prediction algorithm and then partitioned into blocks having N rows and M columns. The blocks are clustered into groups using a clustering algorithm. Next, compression is performed on blocks within a cluster before the compressed blocks are further compressed using arithmetic coding. In some implementations, alphabet creation and correlation is performed on the CFA data before prediction transformation, and color-plane correlation is performed after prediction transformation but before partitioning into blocks. This spatial compression method can be performed efficiently on camera to compress data files, reducing the storage and data flow requirements of the camera.