In this paper, we propose a novel change detection technique using land use patterns as a reference. Detection process is a combination of post classification and algebra techniques, i.e. each of bi-temporal images will be segmented and applying two kinds of algebraic operation on the images covers addition and substraction of two images. Subsequently, we will classifiy each of pixel into four classes namely class 1 and class 4 (related to uninteresting areas), class 2 (associated with no-change area), and class 3 (the growth of new areas in accordance with pattern which was identified from the second image). The pattern for class 3 is identified using classification model that is resulted from the previous research. The model will classify the area into corresponding patterns, namely corridor, block, cubicles, network, geometric, holes, or diffuse. The result of this technique is able to differentiate the change or no-change area.