Segmentation of finger vein image based on pattern feature amplification

To solve the problem of pattern extraction from infrared finger vein images, a new concept of small signal amplification of images based on its pattern feature is proposed. Due to transformation on images, the gradient pattern of boundaries has sparseness in transformation domain. Then the gradient pattern can be preserved and amplified at multi-level transformation domain. So the boundaries will be reconstructed and provide important information for following segmentation by inverse transformation. Secondly, an active contour segmentation model is give which based on prior shapes. At last, the concept of image´s pattern feature amplification is preliminary identified by experimental result. Meanwhile the problem for gray level distortion and local convergence of segmentation curve is solved.