Research On Application Of Cuckoo Search Algorithm In Image Processing

Because of the complicacy of images, the traditional techniques for image processing and analysis have very high computational complexity. As as result, the image processing techniques based on evolutionary computation have been attracted much attentation. Cuckoo search is a new meta-heuristics algorithm, which is better than genetic algorithm and particle swarm optimization in the aspect of ability of global search; it has an applicable potential in image processing. Improved cuckoo search algorithm integrates the advantages of cuckoo search algorithm and particle swarm optimization algorithm, which has strong global optimization ability and good application potential in image processing. In this paper, the improved cuckoo search algorithm is applied into image enhancement, image thresholding segmentation, texture feature extraction and image classification. The main work is as follows:1. An image enhancement method based on improved cuckoo search is implemented, which adopts incomplete Beta function proposed by Tubbs, and makes gray level transformation for degraded image to automaticly search the optimal parameter combination of incomplete Beta function. The proposed method is conducted on some images. Experimental results show that the proposed method is suitable for a varieity of degraded images, and has superior performance for dark and light images.2. An image segmentation method based on improved cuckoo search is designed, which utilizes the global optimization ability of cuckoo search algorithm, optimizing some commonly used image segmentation method, like 2-D Fisher criterion, 2-D maximum Kapur entropy, 2-D minimum cross entropy and maximum fuzzy entropy(single threshold, multi-thresholds). Experimental results prove that improved cuckoo search can quickly obtain the optimal threshold, and significantly reduce the computing time of basic segmentation method.3. A novel method to learn texture mask based on improved cuckoo search algorithm is presented, which uses improved cuckoo search algorithm to train “Tuned” texture mask proposed by J.You. Experimental results demonstrate that the proposed method can adaptively obtain the robust texture mask, and the mask has good discrimination accuracy for typical image texture.4. An image classification method based on improved cuckoo search and support vector machine(SVM) is put forward. For image classification, we usually extract high-dimensional image feature before classification. As a consequence, in the paper, when SVM is applied for image classification; firstly, we adopt improved cuckoo search for optimal feature subset selection, and choose the feature sunset which is able to represent the essential features of image samples; further, we use SVM for image classification, and optimize the parameters of SVM. Experimental results demonstrate that the proposed method can improve the efficiency and accuracy of image classification.On the whole, we have carried on research of image enhancement, image segmentation, texture feature extraction and image classification based on improved cuckoo search, experimental results indicate that improved cuckoo search has good feasibility and effectiveness for image processing and analysis, and has a wide application prospect in this field.