科学技术与工程
科學技術與工程
과학기술여공정
Science Technology and Engineering
2015年
32期
175-180
,共6页
自动对焦%爬山法%清晰度评价值%高清摄像机
自動對焦%爬山法%清晰度評價值%高清攝像機
자동대초%파산법%청석도평개치%고청섭상궤
auto-focus%climbing algorithm%focal value%HD camera
为了改善经典爬山法的性能同时进行算法综合性能评价,提出基于双向两次下降的自动对焦算法和一种综合性能评价指标。首先,根据清晰度评价值的大小确定初始对焦方向。接着,以变化步长方式搜索,直至出现清晰度评价值下降。然后,按照双向两次下降法确定是否处于评价函数波峰的下降位置。最后,以单步长方式在检测到的波峰下降位置处搜索波峰位置即为对焦目标位置。在用18倍数字高清摄像机采集的两个图像序列中比较了本文方法和几种经典的对焦算法的性能差别,实验结果表明:提出的方法在平均过冲步数和平均行程步数这两个指标上具有明显优势,并且在其他分项评价指标上没有明显的短板,其综合性能评价指标为0.98,高于爬山法、改进爬山法、全局搜索法和二分搜索法等算法的综合性能评价指标值。
為瞭改善經典爬山法的性能同時進行算法綜閤性能評價,提齣基于雙嚮兩次下降的自動對焦算法和一種綜閤性能評價指標。首先,根據清晰度評價值的大小確定初始對焦方嚮。接著,以變化步長方式搜索,直至齣現清晰度評價值下降。然後,按照雙嚮兩次下降法確定是否處于評價函數波峰的下降位置。最後,以單步長方式在檢測到的波峰下降位置處搜索波峰位置即為對焦目標位置。在用18倍數字高清攝像機採集的兩箇圖像序列中比較瞭本文方法和幾種經典的對焦算法的性能差彆,實驗結果錶明:提齣的方法在平均過遲步數和平均行程步數這兩箇指標上具有明顯優勢,併且在其他分項評價指標上沒有明顯的短闆,其綜閤性能評價指標為0.98,高于爬山法、改進爬山法、全跼搜索法和二分搜索法等算法的綜閤性能評價指標值。
위료개선경전파산법적성능동시진행산법종합성능평개,제출기우쌍향량차하강적자동대초산법화일충종합성능평개지표。수선,근거청석도평개치적대소학정초시대초방향。접착,이변화보장방식수색,직지출현청석도평개치하강。연후,안조쌍향량차하강법학정시부처우평개함수파봉적하강위치。최후,이단보장방식재검측도적파봉하강위치처수색파봉위치즉위대초목표위치。재용18배수자고청섭상궤채집적량개도상서렬중비교료본문방법화궤충경전적대초산법적성능차별,실험결과표명:제출적방법재평균과충보수화평균행정보수저량개지표상구유명현우세,병차재기타분항평개지표상몰유명현적단판,기종합성능평개지표위0.98,고우파산법、개진파산법、전국수색법화이분수색법등산법적종합성능평개지표치。
In order to improve the performance of classic climbing method while evaluating the performance of the new method.A robust auto-focus algorithm and a comprehensive performace evaluation method are proposed. First, the direction of the initial searching is determined by focal values on the initial focus position and its adjacent position.Then,a searching stratage with varialbe steps is executed until focal value decreases.Then, the two-way twice descent method is carried out to determine whether it is in a downward position of focal value function peak. Finally, from the downward position of focal value function, a single-step searching method is used for searching the focused position.Comparison of the proposed method and some classical methods used in two image sequences cap-tured by 18 x digital high-definition cameras is demonstrated.The experimental results showed that:this method sig-nificantly reduces the in-focus position overturn steps and total moved distance.In addition, there is no obvious shortcomings on the other sub-index in this new auto-focus method.The comprehensive performance evaluation in-dex of the new auto-focus method is 0.98, and is higher than all the other methods.