Efficient auto-focus algorithm utilizing discrete difference equation prediction model for digital still cameras

doi:10.1109/TCE.2006.273125

2006-11-01

PRE

Chih‐Ming Chen *

Chin‐Ming Hong

Han-Chun Chuang

DOI：10.1109/TCE.2006.273125

原文链接

原文求助

摘要

En 中文

Most digital still cameras employ the passive auto-focus scheme to achieve these goals. Measuring the sharpness of an image using passive auto-focus is time-consuming, particularly for digital still cameras with mega-pixels resolutions. However, the auto-focus search algorithm for digital still cameras needs to be fast and real-time. This study presents an efficient real-time auto-focus (AF) algorithm combining the discrete difference equation prediction model (DDEPM) and bisection. search method for digital still cameras. The proposed auto-focus algorithm uses both the coarse search procedure with DDEPM and the fine search procedure with bisection search method to perform the auto-focus process for digital still cameras. The coarse search procedure utilizes the precise forecasting ability of the DDEPM model to determine the turning points of the focus value curve, thus skipping some unnecessary detection points on the focus value curve. This procedure can be used to move the lens quickly to the neighborhood region of the best-focused lens position during AF processes. After the coarse search is finished, the fine search procedure is used to find the best-focused lens position according to the coarse search result. The proposed scheme increases the auto-focus speed since it reduces the number of detection points required for the focus value curve when searching the neighboring region of the best-focused lens position, and also reduces the lens backlash times. The proposed algorithm is implemented herein on an HP digital camera including control software. The experimental results indicate that the proposed AF algorithm is very reliable and requires few focusing iterations.

Keyword:

auto-focus

digital still cameras

discrete difference equation prediction model (DDEPM)