<p>Image contrast enhancement is a common area under preprocessing. The type of enhancement to be addressed reflects upon the type of image degradation present. However, the enhancement along with information preservation is a challenge. Many recent deep learning (DL) work has been done towards image enhancement but a few stresses both. It is often seen that most methods shift the pixels present in low-intensity regions to the higher ones. This improves the contrast but no conclusion about its preserved features can be made. To address this issue, this work uses increased skip connections. The skip connections are modified in a way that uses attention blocks and sharpen filter. The convolution block attention module (CBAM) is used to preserve the important features of the input image. Sharpen filter is used in the customised skip connections of the encoder block to avoid the blurring of image features due to the spatial gap between two convolution layers in the encoder. Three different types of datasets are used. The aerial image dataset taken from the university of south california, signal and image processing institute (USC-SIPI), sentinal-2 dataset, and the landsat satellite imagery. For the US-SIPI dataset, PMSIRE is used for generating ground truth data. For sentinal and landsat, the original images are used as masks and the low brightness images are generated using the controlled transformations. It has been observed that the average discrete entropy (DE) of the proposed method is 5.8211, The average contrast per pixel (CPP) is 21.93, measurement of enhnacement (EME) is 14.0898, structural similarity index measure (SSIM) is 0.9247, BRISQUE is 12.5049, and time is 1.596&#xa0;seconds, which are superior to nine conventional methods. Therefore, this method can be a forementioned practice in the field of satellite imagery. The code is released on <a href="https://github.com/P321v/Enhancement1">https://github.com/P321v/Enhancement1</a></p>