採用對比調整與曝光控制之單鏡頭高動態視訊技術

Abstract

近日內，隨著科技快速發展，人類追求更為真實的視覺體驗，接近真實場景的影像越顯得重要。因此，在各種視訊應用當中，可以提供更廣動態範圍及更佳影像品質的高動態範圍(HDR)影像技術吸引了人們的注意。 現今有不少專門研究單鏡頭高動態範圍技術的文章，而這些需要曝光良好輸入影像的方法在多數極端場景中(如：同時存有高亮度及低亮度的場景)，仍無法讓亮處或暗處的細節資訊在現有低動態範圍的顯示器上呈現出來。 本論文提出一個單張影像具曝光控制的HDR影像增強技術來解決極端(如強烈前光及背光等)場景所造成之細節無法呈現的問題。本論文所提出的系統分成兩部分：第一，整合動態區域對比度增強(Dynamic Local Contrast Enhancement, DLCE)技術及模糊增強(Fuzzy enhancement)技術的單張影像HDR成像技術，能呈現同時具有高亮度及低亮度內容的場景。第二，利用區域平均、相機反應函數及主要物件偵測技術所發展出來的曝光控制方法，可得到一張良好曝光的影像。 最後，我們使用TI OMAP4430應用處理器(具雙核心ARM Cortex-A9處理核心)來進行本論文所提方法之嵌入式系統實作，經過優化後可達到每秒約13張VGA畫面之處理效能。

With the rapid development of modern technology, human beings pursuit of more realistic visual experiences. Image quality closing to the real world scenery is getting more and more important. Therefore, the High Dynamic Range (HDR) imaging technique, which provides a wider dynamic range image and results in a better quality image, engages people’s growing attention in a variety of applications. There have been many single-camera HDR imaging papers published in the literatures. However, those methods needing appropriate exposure control still cannot reveal details of the bright part and dark part existed at the same time in many extreme conditions such as scenes with both high luminance and low luminance parts. Moreover, when there is an excessive front lit in the scene, all of the approaches are not capable of handling the over-exposed parts. In this thesis, a single-camera HDR imaging algorithm, which is able to solve all the extreme conditions by controlling the exposure time to get a well-exposed image and enhance image quality, is proposed. The proposed system is divided into two parts: i.e. single-image HDR imaging technique and an exposure control method. In the single-image HDR imaging technique, we integrate a dynamic local contrast enhancement (DLCE) method and fuzzy enhancement, which can handle both glaring regions and dark regions existed at the same scene. In the exposure control method, we adopt local average, camera response function and main object detection to achieve a well-exposed image. Finally, we adopt the TI OMAP4430 to implement the proposed system on an embedded environment with the performance about VGA video@13 frames per second.