應用形態學於點對點等比例縮放影像之方法

Apply Morphology to Pixel-by-pixel Proportionate Static Binary-Image Enlargement

蘇怡禎(I-Chen Su)

指導教授 ： 江行全

元智大學/工程學院/工業工程與管理學系/碩士(2009年)

https://doi.org/10.6838/YZU.2009.00122

摘要

影像放大是數位影像處理中常用到的功能，因此本研究透過改良影像處理的形態學(Morphology)，將靜態影像(static image enlargement)點對點(Pixel by pixel)等比例放大。
傳統上，膨脹(Dilation)、侵蝕(Erosion)影像後，往往會喪失轉角點資訊，本研究針對不同角度、不同方向之轉角點自動給予轉角點不同大小之遮罩，透過形態學(Morphology)縮放影像後仍能保持邊角資訊，將二值影像(Binary image) 點對點(Pixel by pixel)等比例縮放。第一部份將原始影像透過Otsu強化，第二部份邊緣偵測部份採用邊緣追蹤法(Boundary following)偵測脹縮後影像及原始影像之邊緣，第三部份則透過物件輪廓點估算每一個邊緣點之真圓度(Roundness)，將真圓度與閥值比較以偵測轉角點，並計算轉角點之角度與方向，第四部份是本研究重點，進行轉角點修補，透過膨脹後之邊緣計算轉角點所需之遮罩大小，紀錄二值化後之原始影像轉角點資訊自動針對脹縮後影像之轉角點進行補角。
本研究目前針對銳角轉角點間距80度至10度進行程式敏感度測試，當於轉角點間距大於30度之測試影像能有效修補不同方向之轉角點放大率為100%。

關鍵字

形態學、邊緣追蹤法、邊緣偵測、轉角點偵測、真圓度、影像縮放

並列摘要

Image erosion and dilation are common functions in digital image processing. This study aims to improve the morphology quality by pixel-by-pixel processing for an image enlargement or reduction.
Traditionally, an image tends to lose its preciseness after applying morphology. This study automatically supplies masks of various sizes to the corners of different angles and orientations; image enlargement/reduction through morphology will then be able to retain its corner information and to enlarge/reduce the gray-level image pixel-by-pixel. The first phase involves in binarizing an image using the Otsu method, while the second phase utilizes the boundary following technique to detect the boundaries of the dilated image and the original image. The third phase estimates the roundness of each edge point via the contour points of the object; the roundness and threshold are then compared in order to detect the corner points and calculate their angle and orientation. The fourth phase is crucial to this research; using corner point inpainting, the dimension of the mask required by each corner point is calculated through the dilation/erosion of the image boundary. The corner point information of the original image is binarized and recorded, then inpainting is performed to the corner points of the enlarged/reduced image.

並列關鍵字

Morphology, boundary following, boundary detection, corner detection, roundness, image enlargement

參考文獻

A. Bovik and J. Gibson. (2000), Handbook of Image and Video Processing, Academic Press, 1nd Edtion ,New York, USA

C. T. Lin, K. W. Fan, H. C. Pu, S. M. Lu, and S. F. Liang. (2007), “An HVS-Directed Neural-Network-Based Image Resolution Enhancement Scheme for Image Resizing” , IEEE Transaction on Fuzzy Systems, Vol. 15, No. 4, pp.605-615.

C.H. Hsieh and Y.J. Liu. (2000), “Fast search algorithms for vector quantization of images using multiple triangle inequalities and wavelet transform” , IEEE Transactions Image Processing, Vol. 9, No. 3, pp. 321-328.

C. H. Hsieh, J. C. Tsai, and P. C. Lu. (1996), “Noiseless coding of VQ index using index grouping algorithm” ,IEEE Transactions Communications, Vol. 44, No. 12, pp. 1643-1648.

R. C. Gonzalez and R. E. Woods.(2002), Digital Image Processing, Pearson Education,2nd Edtion, New Jersey, USA.