標題: 基於失真形式偵測的影像品質評估方法
Hybrid Image Quality Assessment Method Based On Distortion Type Detection
作者: 郭鴻一
Kuo, Hung-Yi
蔡文錦
Tsai, Wen-Jiin
多媒體工程研究所
關鍵字: 影像品質評估;結構相似性指標;考量人類視覺系統之峰值信噪比;圖像失真形式偵測;Image Quality Assessment (IQA);Structural Similarity Index (SSIM);Peak Signal to Noise Ratio-HVS (PSNR-HVS);Image Distortion Type Detection
公開日期: 2014
摘要: 影像品質評估(Image Quality Assessment, IQA)在許多與數位影像有關之應用上都是相當重要的問題,所以目前已有許多方法被提出,以僅可能正確的估量影像的品質為目的。然而,過往的實驗數據顯示,一個影像品質評估方法可能在某些特定形式的圖像失真(Distortion)上有較佳的表現,卻在其他的失真形式上表現較差。 本論文提出一種直觀的方法來偵測圖像的失真形式,並且依偵測的結果自兩種目前常用的視覺評估方法-考量人類視覺系統之峰值信噪比(PSNR-HVS)和結構相似性指標(SSIM)之中,選用其較適合者。實驗結果表明本方法可確實的結合兩方法之優點,並且相較於兩者在圖像的品質評估上皆有較好的表現。
Image Quality Assessment is an important issue on many applications about digital image, so there are many methods were proposed to evaluate quality of an image as correctly as possible. However, recent experiment results show that an IQA metric may perform well on some distortion types of images, but bad on other distortion types. In this paper, we propose an intuitive method to detect the distortion type of an image, and use the detection result to choose appropriate metric from two widely used IQA methods : PSNR-HVS & SSIM. Our experimental results show that our method performs better than PSNR-HVS and SSIM
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT070156634
http://hdl.handle.net/11536/75945
顯示於類別:畢業論文