Abstract
Object recognition is one of the most fundamental issues of image processing. Due to the significant variations exhibited by the diversified real world patterns, it is a challenging issue against the inconsistent illumination, partial occlusion, changing background and shifting viewpoint. The merits of image fusion lie in its reliability and capability for object recognition in terms of actual redundancy and complementary information. However, information loss and aliasing impairment might occur. It is essential that image information should be adequately aligned and registered ahead of combining the images. For the 2D wavelet technique, discrete wavelet transform is used to decompose images and reconstruct source images using components of approximation, horizontal detail, vertical detail and diagonal detail. On the other hand, images can be enhanced by contrast stretching from the human visual perspective. Adaptive contrast stretching adapts to intensity distributions to improve image quality under poor visual appeals and illumination conditions. It provides sensitive information through the sharper contrast changes and hidden feature indications, while the possible noise amplifying can be resolved by adaptive algorithms together with bilinear interpolations. The limit constraint enhancement should be conducted against the over-saturation. For similar images, unique patterns occur at each individual. Integration of image fusion and adaptive contrast stretching is thus proposed to enhance the quality of object recognition. The quality of image processing techniques should be judged by not only the visual perspectives but also the quantitative measures. Both subjective and objective assessments should be conducted. Thus, the information theory has been employed, where notions of the discrete entropy, gray level energy, mutual information, relative entropy and information redundancy have been proposed to illustrate the impact of image processing integration of 2D wavelet fusion and adaptive histogram equalization techniques. This approach can be easily expanded to a wide variety of other image processing issues.

Keywords: Wavelet Fusion, Adaptive Contrast Stretching, Discrete Entropy, Relative Entropy, Gray Level Energy, Mutual Information, Information Redundancy.

(P1150837362, 2.77 MB)

BibTex:

@ARTICLE{P1150837362,

AUTHOR = {Zhengmao Ye and Yongmao Ye and  Yin Hang and Habib Mohamadian},

TITLE = {Integration of Wavelet Fusion and Adaptive Contrast Stretching for Object Recognition with Quantitative Information Assessment},

JOURNAL ={ICGST International Journal on Graphics, Vision and Image Processing, GVIP},

}

(P1150837362, 2.77 MB)