Abstract :
Adaptive decimation (AD) is a technique that aims at compressing images with a very small amount of computations and memory requirement. For images that contain moderate amounts of textural content, the method exhibits satisfactory performance and in general provides good visual quality and acceptable coding fidelity at low bit-rate of around 0.2 bpp. Although the complexity of the method is relatively light when compared with the existing compression methods, it still involves a considerable amount of computations that would require the use of medium-speed processors to achieve real-time operation. In the paper, a novel image encoder based on the principles of AD is reported. The scheme is near-computation-free as it involves on average a single fixed point multiplication plus a few other summing and logical operations for every four pixels. Experimental results reveal that, despite the substantial reduction in complexity, the performance of proposed method is similar to, if not better than, the existing AD encoding algorithms.
