1-bit heuristic adaptive quantizer (HAQ) for on chip image compression in CMOS image sensors

This paper presents an algorithm for implementing a single bit adaptive quantizer based on fast boundary adaptation rule (FBAR). The peak signal to noise ratio (PSNR) gain and performance gain of the algorithm over prior designs is found to be larger than that displayed by prior art compared with a reference FBAR implementation. A maximum increase of 1.44db was seen. In addition, the new design facilitates an improved bits per pixel ratio (bpp) when integrated with the QTD compressor utilized in previous prototypes. The presented algorithm is hardware friendly and designed for low power implementation, with simulation results also showing an improvement of relative energy cost over previous work. Experimental evidence for image sizes ranging from 64×64 pixels to 512×512 pixels and the heuristic adaptive quantizer (HAQ) algorithm are detailed in this paper.