Predictive differential modulation for CFA compression
Arcangelo Ranieri Bruna, Filippo Vella, Antonio Buemi, Salvatore Curti · Nordic Signal Processing Symposium · 2004
The recent, wide diffusion of Digital Still Cameras (DSCs) and mobile imaging devices disposes the need of developing efficient techniques for digital image coding, in order to reduce resources requirements for their storing and transmission. Digital cameras usually acquire the image data through a Bayer Color Filter Array (CFA), a light sensitive sensor able to acquire the red, green and blue colors. Each pixel acquires just one color component. The full color image is then restored applying a sequence of image processing algorithms interpolating the acquired data. This paper introduces a new, efficient coding method to compress the Bayer Pattern. It is based on a predictive schema followed by a Vector Quantization (VQ) technique. Simulations have demonstrated that the proposed scheme allows a visually lossless compression of Bayer pattern images with low memory cost.