Breast cancer (BC) is the most common cancer among women world-wide, approximately 20-25% of BCs are HER-2 positive. Analysis of HER-2 is fundamental to defining the appropriate therapy for patients with breast cancer. Inter-pathologist variability in the test results can affect diagnostic accuracy. The present study intends to propose an automatic scoring HER-2 algorithm. Based on color features, the technique is fully-automated and avoids segmentation, showing a concordance higher than 90% with a pathologist in the experiments realized.