| Summary: | 碩士 === 國立中央大學 === 資訊工程學系在職專班 === 101 === Webcams are actively used for social interaction. Accordingly, this sort of application has been successfully visualized in many consumer electronics products, such as notebooks, tablets and smart TVs. In the context of user-driven processing operations, focusing is probably the most frequently used function developed to obtain a particular desired view of a scene. In the case of fixed-focusing webcams, the captured scene will exhibit blurring contents if the image scene is out of focus. It hence drives the demands for achieving a particular desired clarity in the scene with an auto-focusing function.
Almost all of the auto-focusing webcams in the market are based on a system-on-chip(SOC) framework, where a integrated camera module includes lens, complementary metal-oxide-semiconductor(CMOS) sensor, image signal processor(ISP) and voice coil motor(VCM). Due to the consideration of cost, many different auto-focusing approaches have been developed to overcome the limitations pertaining to resources available to consumer webcams. Among these, hill-climbing algorithm is most widely adopted because of its simplicity and easy implementation in hardware for real-time applications. However, traditional hill-climbing solution is limited to focusing on high-frequency blocks of an image scene. This results in a miss-focusing when the complex of region-of-interest (ROI) is much simpler than other contents of an image scene.
Motivated by this, we propose an improved hill-climbing algorithm using an adaptive weighting estimation mechanism in this thesis. Experimental results show that this new solution can perform outstandingly well for various scenes and different light conditions without requiring high computational cost. Moreover, the developed scheme is well-suited for implementation in low-cost consumer webcams.
|
|---|
