Mine image enhancement algorithm based on nonsubsampled contourlet transform
-
-
Abstract
To improve the observability mine images with the low illumination and high noise,a mine image enhancement algorithm based on non-subsampled contourlet transform is proposed. The algorithm overcomes the shortcomings of conventional image enhancement algorithms that cannot take into account both contrast enhancement and noise suppression. In the paper,an image enhancement framework based on Retinex for low illumination image with noise is deduced. The framework removes the interference of noise to the estimated illumination images,and separates the contrast enhancement and noise suppression of images. According to the framework,firstly,the input image is decomposed into low-frequency sub-band coefficients and high frequency directional sub-band coefficients by using subsampled contourlet transform,thus the coupling between esti-mating illumination image and suppressing noise is removed. Secondly,in the contourlet transform domain,the bright channel image of R,G and B channel is calculated using low frequency sub-band coefficients of three channels. However,the characteristics of details mutation and too low gray values of the bright channel image do not accord with the characteristics of slow change of illumination image. Thus,through further Gamma correction and mean filtering of the bright channel image,the estimated value of the smoothed illumination map with improved gray value is obtained. After that,in the contourlet transform domain,the high frequency direction subband coefficients are shrunk according to the threshold function to achieve noise suppression. At last,to highlight the details of a frequency band direction,the direction sub-band coefficients multiplied by the corresponding gain are used to enhance the details. The mine image with noise suppression and detail enhancement is reconstructed by using low frequency sub-band coefficients and high frequency directional sub-band coefficients with detail enhancement. In order to further improve the contrast,the reconstructed image is divided by the estimated illumination image to obtain the final enhanced image. To highlight the details of a certain band direction and improve the overall contrast,the contracted high frequency direction sub-band co-efficients are multiplied by the corresponding gain to complete the specific detail enhancement. The enhanced high frequency sub-band coefficients,the low-frequency sub-band coefficients and the estimated illumination map are used to reconstruct an enhanced image with improved overall contrast. Numerical experiments show that the algorithm can effectively improve the contrast,suppress noise and highlight details. Moreover,the algorithm has a better stability and adaptability,and can well meet the needs of mine image enhancement.
-
-