Abstract. Conventional image processing technology does have limitations in complicated application. An obvious shortcoming of conventional approach is its processing mainly implemented based on pixel by pixel individually, paying more attention to local information and neglecting the global model relations. This presentation deals with two advanced image processing applications: image magnification and face recognition; both were developed based on multiple model features. To magnify an image, in essence, is to add more new pixels between original adjacent pixels in the image. The new pixels are also called interpreting pixels. The proposal advanced magnification approach for generating interpreting pixels is as follows: first, extracting textures of image, decomposing image into sub-regions separated by texture contours; then analyzing and calculating fractal dimension parameters of contours and sub-regions; final, magnifying the texture contours and generating interpreting pixels within each sub-region taking into account of corresponding fractal relations. Face image recognition is an important research topic related to security applications; however, it is also a challenge. There are always variances existing in photos of the same person; this might be caused by variances of lighting conditions, different poses of the subject person, different positions of the camera and other random factors. The pixel-based matching approach does not work validly in this case. A human face can be viewed as a Markov model system, and the morphing photos are observation states of this system. Based on the above principle, a Markov model corresponding to a human face can be built through training operation, and the recognition is based on this model parameters. Some experimental result shows that the Markov-based approach achieved high recognition rate (>90%) and is also robust.