A hybrid adaptive block truncation coding (HABTC) method is presented to improve block truncation coding (BTC)-related compression methods for gray-level images. The basic idea behind the method is to use various coding schemes to take advantage of local image characteristics. A simple linear interpolation coding scheme and a very basic predictive coding scheme are used to improve the compression ratio of the homogeneous area of the image. Moreover, a four-level BTC, whose thresholds are obtained using a radius-weighted mean (RWM), is applied to encode the inhomogeneous area. The bits per pixel/peak SNR (bpp/PSNR) values listed in many other BTC-related papers are cited and compared with ours. It is found that reasonable compression ratio is obtained by the proposed HABTC method and the visual quality of the decoded images is also acceptable.