2026-11-092026-11-09https://scholars.csmu.edu.tw/handle/123456789/38604Recently, the technology of digital capturing device is progressive constantly; it becomes a very important role in people life. For example, the digital camera or digital video is often used wherever work or travel. Additionally, automatic Surveillance System is an enabling technology that is in strong demand by large application sections, such as subway station or airport. Let the image quality captured by any digital capturing device becomes very important. Therefore, the research project will improve auto-exposure function of conventional digital capturing device. Let them have the ability of detection, classification and compensation for input image. This research is motivated by the above mentioned issues. We proposed to develop an intelligence digital image auto-exposure classification and compensation system with DSP realization. The proposed project includes three parts: (1) image feature detection and classification, (2) the compensation of problem image and (3) algorithm implementation with DSP chip. We first extract the image feature from input image captured by any digital capturing device. Each image features can display the characteristic of under-exposure, over-exposure and backlight image. When getting the image features, we start performing image classification task. The objective of image classification is for understanding image category. The preliminary method of image classification is using fuzzy inference method. After image classification, it is to achieve image compensation. We use a cubic compensation curve to adjust image brightness for different category input image. Finally, we will port the development algorithm into a DSP chip so that it can practically apply to commercial product.近年來，由於數位取像裝置的技術不斷的進步，使得它在人類的生活中扮演著很重 要的角色，如數位相機和數位攝影機就是人們常會使用到的，不論是工作或者是旅遊都 常不離手，同時近來自動監控系統大量的使用於特定區域或場所，例如公寓大廈、地鐵 站和飛機場等，使得所拍攝到影像的品質變的非常的重要，因此本研究計畫將針對傳統 數位相機和取像裝置中影像自動曝光的功能做調整改善，使其功能可以針對影像的曝光 不足、曝光過度和逆光作偵測、分類和補償，並將所發展出來的演算法用數位訊號處理 器晶片加以硬體實踐。 故本計劃的研究動機就在於研發智慧型的數位影像自動曝光分類與補償系統的軟 硬體，其內容可以分成三大部分：（1）影像特性的偵測與分類；（2）問題亮度影像的 補償；（3）演算法的硬體實踐。我們首先將輸入的影像作特徵的抽取，每個特徵要都 能表現出曝光不足、過度與逆光的特性，當獲得影像的特徵後便會進行影像的分類工 作，初步所採用的分類方式是模糊推論(Fuzzy Inference)目的是為了清楚輸入影像的類 別，最後一個步驟是進行影像的補償，針對上一個步驟的結果，我們提出利用三次曲線 (Cubic Curve)補償法來提昇影像的亮度和改善影像的品質。當軟體系統研發完成後，本 計畫會將它移植到數位訊後處理器的晶片內，如此系統才能真正的用於實際的商業產品 中。數位訊號處理器自動曝光自動監控系統模糊推論三次曲線曝光不足曝光過度逆光DSP ChipAuto-ExposureAutomatic Surveillance SystemFuzzy InferenceCubic Compensation CurveBacklight ImageProject