Choosing the filter for catenary image enhancement method based on the non-subsampled contourlet transform

The quality of image enhancement plays an important role in the catenary fault diagnosis system based on the image processing technique. It is necessary to enhance the low contrast image of catenary for better detecting the state of catenary part. The Non-subsampled Contourlet transform (NSCT) is the improved Contourlet transform (CT), which can effectively solve the problem of artifact phenomenon in the enhanced catenary image. Besides, choosing the enhancement function and the filter of the NSCT will directly influence the image enhancement effect. In this paper, the proposed method is implemented by combining the NSCT with the nonlinear enhancement function to enhance the catenary image. First, how to choose the filter of the NSCT is discussed. Second, the NSCT is used to decompose the image. Then, the chosen nonlinear enhancement function is used to process the decomposed coefficient of the NSCT. Finally, the NSCT is inversed to obtain the enhanced image. In this paper, we evaluate our algorithm using the lifting wavelet transform, retinex enhancement method, dark channel enhancement method, curvelet transform, and CT method as a comparison to enhance a group of randomly selected low contrast catenary images, respectively. The results of comparative experiments conducted show that the proposed method can effectively enhance the catenary image, the contrast of image is improved, the catenary parts are obvious, and the artifact phenomenon is effectively eliminated, where image details (edges, textures, or smooth areas) are also well preserved. Besides, the values (detail variance-background variance, signal-to-noise ratio, and edge preservation index) of measuring the image enhancement capacity are improved, while the mean squared error value is decreased when compared to the CT method. These indicate that the proposed method is an excellent catenary image enhancement approach. Published by AIP Publishing.