Contrast optimization by region adaptation (COBRA) for grey scale images

• Main Author: Ahmed, Muhammad Mahmood
• Format: Thesis
• Language: English
• Published: 2013
• Subjects: QA76 Computer software
• Online Access: http://umpir.ump.edu.my/id/eprint/37082/1/Contrast%20optimization%20by%20region%20adaptation%20%28COBRA%29%20for%20grey%20scale%20images.pdf; http://umpir.ump.edu.my/id/eprint/37082/

Medical images form a part of real world images which come with a wide variety of contrast and brightness. The acquired images almost invariably require contrast enhancement. Some of the underlying contrast enhancement methods do not produce predictable results. Contemporary contrast enhancement frequently relies on histogram equalization (HE). In practice, HE produces unexpected results. Such, inconsistent results make reliability of HE questionable. As a result HE is unacceptable in sensitive areas like medical field. The situation leads to a detailed analysis of HE, which brings out that foundation of HE is based on density not contrast. As this foundation is unrelated to contrast, resulting contrast changes are unpredictable. As a solution, a novel method based on factors directly related to contrast is proposed. This method separates the image into dark and bright regions. Based on this concept the proposed method named Contrast Optimization by Region Adaptation (COBRA) is developed by optimizing the contrast ratio of separated regions. To achieve this optimization, the whole image is shrunk to a lower scale which provides necessary space for readjustment of contrast ratio. Then the constituents grey levels are raised exponentially to revert back to original scale. This exponential reversion, adjusts the contrast ratio of separated regions to optimum. This contrast optimization fulfills deficiency in real world images. Due to contrast based foundation of the proposed method, the resultant enhancement in similar type of images is consistent. These predictable results, yield high reliability which makes the proposed method trustworthy for critical areas like medical field. Additionally the results reveal that the histogram of the enhanced image represents similarity with the original image. This similarity is measured using DICE and Jaccard methods. Based on the similarity figures, comparative analysis was carried out between the proposed method and HE. The analysis verified that the proposed method achieves excellent results on brain MRIs, additionally it performs well on general medical and common bench mark images.