Block Truncation Coding (BTC) has been considered as a highly efficient compression technique for decades. However, the annoying false contour caused by a low bit rate configuration is a key problem. Two recently proposed halftoning-based BTC, namely the Error-Diffused BTC (EDBTC) and the Ordered-Dither BTC (ODBTC), can significantly reduce the above issue. However, some defects such as blocking effect and halftoning impulses are also accompanying with the reconstructed image. For this, this work organizes an optimization procedure to generate a number of texture-dependent filter sets to solve these problems. As demonstrated in the experimental results, the proposed strategy can significantly improve the image quality. This improvement makes the former halftoning-based BTC schemes as potential candidates for surveillance and computer vision applications.