在本篇論文中,我們將提出一種以資料的結構性與預測準確度做為預測資料的基礎,利用不甚精確的字典紀錄,使其能有效地預測下一筆資料的可能範圍。藉由概略性的模型,使預測的資料儘可能匹配真實資料,以達到提高資料的冗餘性,縮減資料之間的Entropy,進而輔助其它壓縮演算法來提高其壓縮比,達到簡化資料,增進儲存空間的利用率。 We propose an algorithm according to the data sequence and the degree of accuracy of guess to predict bmp type data in lossless data compression in our research. We research the correlation between two pixels. And our flow of prediction can decrease the entropy of data and aid algorithm of compression to increase its compression ratio.