過去已有不少學者提出有關影像對比度增強的改善方法，舉凡從全域性增強到區域性增強，或是從灰階度轉換增強到對比度轉換增強。然而這些既存的影像對比度增強方法中卻存在著三項問題：(1)未說明對比度的增強幅度是否合乎人類視覺特性，(2)未討論單一種對比度增強模型可否能同時滿足自然影像及醫療影像的需求，(3)在影像增強後的影像品質評估上，很少同時將主觀評估和客觀評估的數據拿來做比較，由此驗證其方法是合乎人類視覺的觀察感受。

　　本研究檢驗多張自然影像及醫療影像中所有像素點的主體亮度(ob) 與背景亮度(bg)差異值L，發現自然影像的L/bg的分佈範圍幾乎是分散在數倍於JND(Just Noticeable Distance)標準值，到低於JND標準值的幾分之一。這表示自然影像可以很容易區分平滑、細部及邊緣的三個區域。相反地，乳房X光影像的L/bg的分佈範圍有七成以上的像素點是處於低於JND標準值的幾分之一。這也驗證了人眼很難觀察出乳房X光片中灰階度的微小變化。同時也說明了不同類型的影像應該要有不同的影像增強模型來改善它的對比度。

　　在自然影像的研究中，我們首先以切割直方圖均等化做前處理，讓整張影像的灰階度不要過度集中於小範圍的灰階度，由此使得影像更為清晰。接著，將前處理過的自然影像的像素預先歸類到平滑區域、細部區域，或邊緣區域。一般而言，平滑區域並不適合改變其對比度。所以根據影像觀察者自己主觀想要強調的是細部區域或邊緣區域，以及符合人類視覺的JND準則的二個前提，本研究提出兩種不同JND權重的區域性對比度增強模型。從我們的實驗結果可以明顯看出，經過JND權重的區域性對比度增強模型處理過自然影像有二項特點：(1)平滑區域可以保留原有特質，(2)增強細部及邊緣區域的對比度獲得提升。再者，主觀評估和客觀評估(整體對比度)的數據也充分指出JND權重的區域性對比度增強模型的效果比ACE增強模型的效果佳。

　　在乳房X光片影像的研究中，由於臨床醫師非常注重可疑症兆的外廓形狀及亮度變化。因此本研究首先以區域增長策略找出每一像素的動態前景區域及背景區域的平均亮度。接著再以第三種JND的區域性對比度增強模型來增強乳房X光片影像的對比度，以及使得可疑症兆的形狀的更加突顯。從整體對比度測量(IQM) 數據的比較中，明確顯示本研究所提的增強模型比RBCE增強模型更能改善醫療影像的對比度的品質。

　　In the past, some researchers proposed articles of image enhancement to improve the contrast of an image, for instance, from global contrast enhancement to local contrast enhancement, or from gray level transformation to contrast transformation. However, these proposed image enhancement methods always ignored three problems: (1) Did the effect of these methods satisfy the human visual property? (2) Were these methods in effect with both natural images and medical images? (3) Both subjective and objective evaluations on these image enhancements were seldom given and compared together to strongly support their effectiveness in accordance with human bservations.

　　In this study, after observing the intensity (ob) of the foreground of a pixel and that (bg) of the background of a pixel in several natural images and medical images (mammograms), we find that the distribution of the ratio between the deviation L, which is |ob-bg| , and the intensity (bg) is quite different for natural images and medical images. As to the natural images, the range of L/bg spans broadly from several multiples of JND(Just Noticeable Distance) standard value to few JND standard value. On the contrast, the value of L/bg is less than tiny part of JND standard value for near 70% of pixels among mammograms. This result concisely interprets the difficulty to sense the subtle change between two neighboring pixels’ intensities in a mammogram. It also indicates that different kinds of image have their distinctive ways on image enhancement to improve their contrast.

　　In the case of natural images, the algorithm of cutting histogram equalization is first applied on an image to deconcentrate its histogram so as to attain a clearer image. Next, pixels of the preprocessed image are allocated into one of smoothed, detail and edge regions. In general, there is no need to adjust the contrast of pixels belonged to the smoothed region. Therefore, in this study, two JND-weight local contrast enhancement models are proposed to meet two key concerns. One is the focus of image viewers on either detail region or edge region, and the other is JND criterion of human visual system. Our experimental results apparently demonstrate that our proposed method could preserve the original property of the smoothed region and reinforce the contrast of these detail region and edge region. Moreover, both the subjective and objective evaluations on our experiments evidently figure that our two JND-weight local contrast enhancement models lead better image quality measures than that of Adaptive Contrast Enhancement (ACE) model.

　　In the case of mammograms, clinic physicians are very cautious with the shape profile and the variety of brightness of a suspicious region. Thus, in this study, the strategy of region growing is first applied to seek the foreground and background region of a pixel in a mammogram, and then their average brightness is determined. Next, the third JND local contrast enhancement model is proposed to increase the contrast of a mammogram and highlight the shape of a symptom. Our experimental results show that our third JND local contrast enhancement could improve the contrast of a mammogram more effective than Region Based Contrast Enhancement (RBCE) because the integrated quality measure of the former model is greater than that of the latter.

[1] 王惠暢, 曾令民, 吳秋文, 雷永耀, 彭芳谷, “乳房疾病新論,” 九州圖書出版, 2000
[2] 洛夫, 林德賽, 郭雅君, 楊雅婷, “乳房聖經,” 郭雅君, 楊雅婷譯, 天下生活出版, 2002
[3] 拜塞特, 戈爾德, 金米史密思, “乳房攝影及品質指南=Work for Quality mammography,” 魯承章譯, 合記圖書出版, 1992
[4] 張金堅, “乳房醫學,” 健康世界雜誌, 1997.
[5] 游溰濤, “乳房健康100問:漫談乳腺癌的預防與治療,” 旺文出版, 1995.
[6] 雷永耀, 彭芳谷, 陳啟明, “乳房疾病=Breast diseases,” 藝軒發行, 1998.
[7] 劉佩綺, 克雷狄恩, “乳癌與妳:防治乳癌必須知道的事,” 克雷狄恩著, 劉佩綺譯, 方智出版, 1996.
[8] A. K. Jain, “Fundamentals of Digital Image Processing,” Englewood Cliffs, NJ: Prentice-Hall, 1989.
[9] A. N. Netravali and B. G. Haskell, “Digital Pictures: Representation and Compress,” New York: Plenum, 1988.
[10] D. -C. Chang and W. -R. Wu, “Image Contrast Enhancement Based on a Histogram Transformation of Local Standard Deviation, IEEE Transactions on Medical Imaging, Vol. 17, No. 4, pp. 518-531, August, 1998.
[11] Chi-Min Liu, Jine-Yi Lin; Kuo-Guan Wu, Chung-Neng Wang, “Objective Image Quality Measure for Block-Based DCT Coding,”IEEE Transactions on Consumer Electronics, Vol. 43, Issue 3, pp. 511-516, 1997.
[12] C. -H. Chou and Y. -C. Li, “A perceptually Tuned Subband Image Coder Based on the Measure of Just-Noticeable-Distortion Profile,” IEEE Trans. on Circuits and System for Video Technology, Vol. 5, No.6, December, 1995.
[13] Rafael C. Gonzalez and Richard E. Woods, “Digital Image Processing,”Second Edition, Prentice Hall, 2002.
[14] J. K. Kim, J. M. Park, K. S. Song and H. W. Park, “Adaptive Mammographic Image Enhancement Using First Derivation and Local Statistics,”IEEE Transactions on Medical Imaging, Vol. 16, No. 5, pp. 495-502, October, 1997.
[15] J. -Y Kim, L. -S. Kim, and S. -H. Hwang, “An Advanced Contrast Enhancement Using Partially Overlapped Sub-Block Histogram Equalization,”IEEE Transactions on Circuits and System for Video Technology, Vol. 11, No. 4, April, 2001.
[16] L. Dash and B. N. Chatterji, “Adaptive contrast enhancement and de-enhancement,”Pattern Recognition, Vol. 24, No. 4, pp. 289-302, 1991.
[17] L. P. Dung, Bala Srinivasan, M. Salahadin, S. Kulkarni, W. Campbell, “A measure for image quality,” ACM Symposium on Applied Computing, pp. 513-519, 1998.
[18] M. D. Levine, “Vision in Man and Machine,” New York: McGraw Hill, 1985.
[19] M. I. Seazn, K.-L. Yip, and S. J. Daly, “Uniform perceptual quantization: Application to digital radiography,” IEEE Trans. Syst. Man Cyber., Vol. SMC-17, pp. 622-634, 1987.
[20] N. Otsu, “A threshold selection method from gray-level histogram,” IEEE Transactions on System Man Cybernetics, Vol. SMC-19, No. 1, pp. 62-66, 1979.
[21] P. -S Liao, T. -S. Chen and P. -C. Chung, “A Fast Algorithm for Multilevel Thresholding,” Journal of Information Science and Engineering, Vol. 17, pp. 713-727, 2001.
[22] P. M. Narendra and R. C. Fitch, “Real-time adaptive contrast enhancement,”IEEE Trans. Pattern Anal. Machine Intell., Vol. PAMI-3, pp. 655-661, June, 1981.
[23] R. J. Safranek and J. D. Johnston, “A perceptually tuned subband image coder with image dependent quantization and post-quantization data compression,” Proc. IEEE Int. Conf., Acoust., Speech, Signal Process., Vol. 3, pp.1945-1948, 1989.
[24] R.H. Gold, L.W. Bassett, “X-ray mammography : History, controversy, and state of art in mammography, thermography and ultrasound in breast cancer detection,” Grune&Straton, 1982.
[25] S. M. Pizer, E. P. Amburn, J. D. Austin, R. Cromartie, A. Geselowitz, T. Greer, B. T. Greer, B. H. Romeny, J. B. Zimmerman, and K. Zuiderveld,“Adaptive Histogram Equalization and its Variations,”Comput. Vision, Graphics and Image Processing, Vol. 39, pp.355-368,1987.
[26] T. –L. Ji, M. K. Sundareshan, and H. Roehrig, “Adaptive Image Contrast Enhancement Based on Human Visual Properties,” IEEE Transactions on Medical Imaging, Vol. 13, No. 4, pp. 573-586, December, 1994.
[27] W. M. Morrow and R. B. Parangape, “Region-Based Contrast Enhancement of Mammograms,”IEEE Transactions on Medical Imaging, Vol. 11, No. 3, pp. 392-406, 1992.
[28] Z. Wang, A.C. Bovik and L. Lu, “Why is Image Quality Assessment so Difficult,”IEEE International Conference on Acoustics, Speech, and Signal Processing, (ICASSP '02 Proceedings), Vol. 4 , pp. IV-3313-IV-3316, 2002.
[29] The Mammographic Image Analysis Society, Digital Mammography Database, http://www.wiau.man.ac.uk/services/MIAS/MIASweb.html, 2003-05-08.
[30] 行政院衛生署網站, “http://www.doh.gov.tw/statistic/index.htm,” 2004/5.