Automatic Focus Techniques And Its Applications In Low Lighting Conditions

With the rapid development of digital image processing techniques, the design of a consumer-level digital camera is inclined to introduce user-friendly products which aim to provide easy way to obtain high quality imaging with minimal user intervention in tasks such as auto-focus. The basic idea of auto-focus is to replace the tedious process of manual focusing with automatic adjustment of the lens of the camera to the right position, ensuring the image is well positioned at the focal plane. Although auto-focus has been deployed in some digital still cameras and used as a standard feature, it is relatively new and necessary technique for digital cameras or camera phones that are equipped with high resolution, high megapixel image sensors. However, current auto-focus techniques face great difficulties in its convergence speed, power consumption, and low lighting conditions.In order to solve these problems in autofocus techniques, this dissertation researched on some theoretical studies both in visible and low lighting conditions, including focusing region selection, contrast measurement, and peak search algorithm. The work and contribution of this research are as follows:1. Focusing region selection based on local entropyFocusing region selection is important for the speed, accuracy, and efficiency of auto-focus. Proper selection of focusing region should take several factors into account, including the motion between imaging device and object of interest, different lighting conditions and etc. Aiming to address the problems in state-of-art methods, an entropy based focusing region selection method was proposed. First, the image was divided into blocks, in which the temporal difference of Laplacian distribution model was calculated respectively. Then, the local entropy of each block was computed and used to find the edge information. Based on the edge information, foreground blocks can thus be obtained and used as the focusing region.2. Contrast measurement based on modified entropyAccurate evaluation of image sharpness is critical for obtaining high quality imaging. Contrast measurement should consider the character of homogeneous region, and adapt to the influence of noise and jitter of imaging device. In order to cope with these problems, a modified entropy based contrast measurement was presented. First, the image was divided into blocks, in which the sharpness was calculated. Then, the sharpness of each block can be obtained by computing the mean value of corresponding three blocks from adjacent lens position.3. Peak search based on modified hill climbing algorithmThe convergence speed and power consumption of auto-focus largely rely on the performance of peak search procedure. It is important and difficult to find the focus position from blurred/noisy image. Current hill climbing algorithms pose great difficulties in its adaptability to the influence of noise and jitter of imaging device. Accordingly, a modified hill climbing algorithm based peak search solution was proposed to enable an ideal focus curve which exhibits a single peak and an absence of plateau.4. Auto-focus techniques in low lighting conditionsA focused image obtained in low light conditions possesses a small contrast value, which may be easily influenced by noise. In this case, contrast measurements may generate fluctuant curves with many local peaks. Accordingly, a comparison among different contrast measurements in passive autofocus systems was presented to investigate their performance in low lighting conditions. And two criteriums were proposed to provide quantitative description of the performance of each contrast measurements. Aiming to address the problems in state-of-art methods, a new passive auto-focus algorithm is proposed to improve the auto-focus performance in low lighting conditions. First, a noise reduction preprocessing is introduced to make the algorithm robust to both additive noise and multiplicative noise. Then, a new contrast measure is presented to bring in local false peaks, ensuring the presence of a well defined focused peak. In order to gauge the performance of the algorithm, a modified peak search algorithm is used in the experiments.