This paper dedicates to study how to construct a perfect high resolution image from images with few different in-focus locals. Multi-focus technology which grabs images focused on different region in the same scene to rebuild a whole-focus image is adopted to perform this work. However, integrating with the advantages of so many images is a considerable difficult task. In view of this, the paper presents an algorithm for fast realization with high precision.
Firstly, a batch of reversible two-dimensional FIR filters is designed to extract useful components from the local focused
images. Based on comprehensive consideration, an efficient calculation flow is presented too. Secondly, a detector (PPE,
point property evaluation) based on gradient energy is designed to evaluate point property of each channel divided by the
FIR filters. According to the evaluation, the components in each channel are synthesized to build corresponding target
channel. Finally, a perfect image can be obtained by adding all the target channels. At last, the algorithm is verified by an
actual application. In this test, one perfect image is reconstructed from 9 local focalized images photographed with
telephoto lens. Besides, other intermediate experiment results also provide support to the ideas adopted in the algorithm.