To find new biomarkers and establish histopathology protein fingerprint models for early detection of oral squamous cell carcinoma (OSCC), laser capture microdissection (LCM) technology was utilized in 21 OSCC tissues and 7 oral leukoplaque (OLK) tissues as well as their adjacent normal tissues. Each sample was then detected by SELDI-TOF-MS technology and CM10 protein chip as well as bioinformatics tools. Three proteomic biomarker patterns were identified. Pattern 1 distinguishes patients with OLK from healthy individuals. Pattern 2 distinguishes patients with OSCC from healthy individuals. Pattern 3 distinguishes patients with OSCC from patients with OLK. The analysis yielded both a specificity and a sensitivity of 90.48% for pattern 1, a specificity of 100.00% and a sensitivity of 85.71% for pattern 2, and a specificity of 100.00% and a sensitivity of 85.71% for pattern 3. Proteome mass/charge 3714, 3515, and 4944 built the distinguished protein peaks between the OSCC tumor and adjacent normal tissues. The accuracy of the blind prediction was 90.48% (38/42). M/Z 15122 and 7569 built the distinguished protein peaks between the OLK and adjacent normal tissues. M/Z 3738 and 11366 built the distinguished protein peaks between the OSCC and the OLK. By employing LCM technology combined with SELDI-TOF-MS technology and bioinformatics approaches, histopathology would not only facilitate the discovery of better biomarkers for OSCC and OLK, but also provide a useful tool for molecular diagnosis by potential biomarker.