基於美學品質重新排列網路圖片

Re-ranking Web Images based on Aesthetic Quality

高紹航(Shao-Hang Kao)

指導教授 ： 鄭卜壬

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.01011

摘要

在這篇論文中，我們提出了將現有圖片搜尋結果基於圖片品質重新排序的方法。較早期的研究中大多專注於圖片搜尋結果與查詢字的相關性，而導致有些在搜尋結果中排序前面的圖片其視覺上的品質太差無法滿足使用者。為了改善此問題，我們利用機器學習的方法從圖片中抽取特徵，並根據其與圖片品質的關係建立回歸模型，並利用此回歸模型之預測值對圖片做排序。排序結果可提升使用者對於搜尋結果的滿意度。除此之外我們也提出了不同於傳統圖片特徵的新特徵，並且經實驗證明能夠有效的提升結果。我們將實驗做在來自Google, Flickr, INRIA, Photo.net以及DPChallenge的圖片上。並且建立了一個線上圖片搜尋系統，用來將Google以及Flickr的搜尋結果重新排序後呈現給使用者。實驗結果顯示我們的方法在美學的標準下的確提升了圖片搜尋的效果。

關鍵字

圖片搜尋 ； 圖片品質 ； 圖片排序

並列摘要

We propose an approach to re-rank images retrieved from existing image search engines based on image quality in aesthetic view. Previous works ranked images mainly based on their relevance to queries. However, it happens that often the top-ranked images cannot satisfy users due to their low visual quality. To present quality images to users, our approach learns a regression model, which combines both conventional and novel image features according to a given quality-image collection. Several experiments have been conducted on the datasets sampled from INRIA Holiday datasets , Photo.net , DPChallenge , Google Image , and Flickr . The experimental results show the feasibility of the proposed approach in searching aesthetically-pleasing Web-image search results for users.

並列關鍵字

Image Search ； Image Quality ； Image Ranking

參考文獻

[1] A. W. M. Smeulders, M. Worring, S. Santini, A. Gupta, and R. Jain. Content-Based Image Retrieval at the End of the Early Years. In IEEE Transactions on Pattern Analysis and Machine Intelligence, 22(12) 1349-1380, 2000.

[2] Canny J.: A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence 8, 6 (1986), 679–698.

[3] Carlo Tomasi, Roberto Manduchi, Bilateral Filtering for Gray and Color Images, proceedings of the ICCV 1998

[6] Che-Hua Yeh, Wai-Seng Ng, Brian A. Barsky and Ming Ouhyoung, An esthetics rule-based ranking system for amateur photos, ACM SIGGRAPH ASIA 2009 Sketches.

[8] Daugman, J.G., "Two-dimensional spectral analysis of cortical receptive field profiles", Vision Res. 20 (10): 847–56