利用遺傳演算法來做適應性對比增強

Genetic Algorithms for Adaptive Contrast Enhancement

郝家銘(Jia-ming Hao)

指導教授 ： 黃朝章

元智大學/工程學院/電機與資訊工程研究所/碩士(1995年)

摘要

遺傳演算法是一種具有高度適應性的搜尋法則，常用來解決最佳化與機器
學習問題。Pal提出使用遺傳演算法來自動搜尋影像增強的灰階轉換函數
。本篇對其提出一改進方案，以四個代替原來十二個搜尋變數，以縮短處
理時間。我們並在適應函數中考量每個像素與鄰近像素的灰階值變化關係
，使圖片灰階的變化更為自然。最後，我們使用三個工具：影像對比值、
影像資訊流失量與局部灰階變異度，來測試對比增強的效率。其目的是希
望在增加影像對比值的同時，能維持低的影像資訊流失量與局灰階變異度
。經由實驗與直方圖均衡法、Pal方法相比較，本方法在視覺效果上與數
據分析上都有良好的表現。

關鍵字

遺傳演算法 ； 對比增強 ； 適應函數

並列摘要

Genetic algorithms (GAs) represent a class of highly parallel
adaptive search processes for solving a wide range of
optimization and machine learning problems. In this paper, a new
technique for contrast enhancement is proposed, which developed
from the method of Pal with its modifications. The suggested
method uses GA to search an appropriate gray-level
transformation function automatically for image contrast
enhancement. To make the algorithm context sensitive, we
consider the relationships between a pixel and the pixels
surrounding it as the fitness function of GA. Finally, we use
three tools to measure the performance of contrast-enhancing
methods. The three measurements are image contrast value, image
information loss value, and local intensity variance value. The
goal is to increase an image''s contrast value while keeping both
the information loss value and the local intensity variance
value low. In the experiments, we have compared the performance
of the suggested method with that of the ordinary histogram
equalization technique and the method of Pal. The results were
then evaluated by the three tools. The suggested method
performed well both analytically and visually.

並列關鍵字

Genetic Algorithms ； Contrast Enhancement ； Fitness function