We propose an algorithm to solve image recognition problems for pattern matching by using stickiness of DNA molecules, and show an applicability of DNA computing in engineering research field. In normal silicon-based computer, this algorithm will not be selected because huge parallel and operation will be needed. There has been a way to extract some features from models for pattern recognition. In this paper, in accordance with the classifications, the features are extracted from Standard Models and all the models in Search Image. The features are applied to nodes in order to construct a network. The nodes of all the models are encoded to DNA sequences. By pouring ligase (enzyme), these single strands are hybridized and ligated each other, and let them double strands. The necessary double strands are sorted with PCR and Gel Electrophoresis in parallel, and detected. Existence of them can determine the position where the same models as the Standard Models exist in Search Image.